IJACSA Volume 14 Issue 12 - thesai.org

IJACSA Volume 14 Issue 12

Copyright Statement: This is an open access publication licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, even commercially as long as the original work is properly cited.

View Full Issue

Paper 1: Computer Vision-based Efficient Segmentation Method for Left Ventricular Epicardium and Endocardium using Deep Learning

Abstract: Segmentation of the Left Ventricular Epicardium and Endocardium remains challenging and significant for valuable investigation of cardiac image classification. Previous research methods did not consider the flexibility of the heart area, so measurements needed to be more consistent and accurate. In addition, previous methods ignored the presence of affectability and additional parts, such as the lung organ inside the frame, during segmentation. Deep learning architectures, specifically convolutional neural networks, have become the primary choice for assessing cardiac medical images. In this context, a Convolutional Neural Network (CNN) can be an effective way to segment the left ventricular epicardium and endocardium as CNN can take data pictures, move enormity to various centers or objects in the image and have the choice to separate one from the other. This research proposes an efficient method for segmenting the left ventricular epicardium and endocardium using the InceptionV3 convolutional neural network. Rather than including fully connected layers on the head of the component maps, the proposed method considers the average of each element map, and the subsequent vector was taken care of legitimately into the SoftMax layer. Data augmentation technique was used to validate the proposed method on large number of dataset images. Besides, the proposed method was validated in publicly available MRI cardiac image datasets. Comprehensive experimental analysis was done by analyzing a large number of performance metrics, i.e., cosine similarity, log cos error, mean absolute error, mean absolute percentage error, mean squared error, mean squared logarithmic error, and root mean squared error. The proposed method depicted superior performance for localization of the left ventricular epicardium and endocardium in terms of all these performance metrics. In addition, the proposed method performed efficiently to get smooth curve for covering the region due to usage of interpolation technique to draw the curve, which made it smoother compared with previous research.

Author 1: A F M Saifuddin Saif

Author 2: Trung Duong

Author 3: Zachary Holden

Keywords: Convolutional neural network; segmentation; computer vision; deep learning

Paper 2: Predicting Alzheimer's Progression in Mild Cognitive Impairment: Longitudinal MRI with HMMs and SVM Classifiers

Abstract: The number of elderly people has increased due to the huge growth in human life expectancy over the past few decades. As a result, age-related illnesses and ailments have become more prevalent, including Alzheimer's Disease (AD). A notable deterioration in cognitive functions, particularly memory and thinking skills, characterizes Mild Cognitive Impairment (MCI), a condition that lies in the middle of normal aging and dementia. Therefore, MCI carries a noticeably higher chance of developing into AD and frequently serves as a prelude to dementia. However, using cutting-edge image processing and machine learning techniques, it is possible to examine and find underlying patterns in these complex diseases. By using these techniques, it is possible to separate groups, identify the causes of such separation, and create disease prediction models. Clinical trials, mostly using cross-sectional Magnetic Resonance Imaging (MRI) data, have extensively looked into the use of MRI for the early identification of AD and MCI. On the other hand, longitudinal studies follow the same subjects over an extended period, giving researchers the chance to investigate cross-sectional trends as well as the development of the disease. Three different techniques are put forth in this study for the analysis and assessment of the structural data found in longitudinal MRI scans. Without considering any other diagnostic measures, this information is used to forecast the progression of those who have been diagnosed with MCI. These techniques utilize Hidden Markov Models (HMMs), which capitalize on the advantages of Support Vector Machine (SVM) classifiers.

Author 1: Deep Himmatbhai Ajabani

Keywords: Alzheimer's disease; image processing; Magnetic Resonance Imaging; Mild Cognitive Impairment; machine learning

Paper 3: Integrating Social Media Data and Historical Stock Prices for Predictive Analysis: A Reinforcement Learning Approach

Abstract: The reliance on data collection for assessing individual behavior and actions has intensified, particularly with the proliferation of digital platforms. People often use the Internet to express their opinions and experiences about various products and services on social media and personal websites. Concurrently, the stock market, a key driver of commercial and industrial growth, has seen a surge in research focused on predicting market trends. The vast array of information on social media regarding public sentiment towards current events, coupled with the known impact of financial news on stock prices, has led to the application of data mining techniques for understanding market volatility. This research proposes a novel method that integrates social media data, encompassing public sentiment, news, and historical stock prices, to predict future stock trends. The approach involves two primary phases. The first phase develops a sentiment analysis (SA) model using three dilated convolution layers for feature extraction and classification. Addressing the challenge of unbalanced classification, a reinforcement learning (RL)-based strategy is employed, wherein an agent receives varied rewards for accurate classification, with a bias towards the minority class. Additionally, a unique clustering-based mutation operator within a differential equation (DE) framework is introduced to initiate the backpropagation (BP) process. The second phase incorporates an attention-based long short-term memory (LSTM) model, merging historical stock prices with sentiment data. An experimental analysis of the study dataset is conducted to determine optimal values for significant parameters, including the reward function.

Author 1: Mei Li

Author 2: Ye Zhang

Keywords: Social media; stock market; sentiment analysis; unbalanced classification; reinforcement learning; differential equation; long short-term memory

Paper 4: Cloud Migration: Identifying the Sources of Potential Technical Challenges and Issues

Abstract: Digital Transformation is emerging as a crucial factor for successful adaptation to the modern digital world for all possible economic and social entities. In recent years, cloud migration, cloud services and computing solutions adoption have been popular enablers for the Digital Transformation. During the Digital Transformation process, organizations and institutions face various technical challenges and implementation problems. This article explores the issues related to cloud migration and existing cloud service models. It investigates the advantages and disadvantages of the most popular cloud services offered by leading service providers, summarizes the main challenges in cloud migration processes, and how organizations can overcome them. Results help organizations understand the sources of potential technical challenges and implementation problems affecting cloud adoption and address these issues at an early stage of the initiative in order to reduce the threat of failure, avoid potential pitfalls and achieve desired cloud capabilities and business benefits.

Author 1: Nevena Staevsky

Author 2: Silvia Gaftandzhieva

Keywords: Digital transformation; cloud; cloud migration; cloud models; PaaS; SaaS; IaaS; challenges

Paper 5: Technology-Mediated Interventions for Autism Spectrum Disorder

Abstract: According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), Autism Spectrum Disorder (ASD) is a complex neurological and developmental condition characterized by impairments in social interaction and communication. Despite significant advancements in the research field, no pharmaceutical medication has been designed for ASD treatment. Therefore, ASD treatment relies mainly on therapeutic intervention. Interactive technologies have emerged as valuable therapy augmentation tools. This research focuses on interactive technologies developed for ASD therapeutic intervention. The study introduces a conceptual framework for understanding the full spectrum of technologies involved in the ASD context. The employed methodology encompasses expert opinions and entails a cross-sectional study that included 59 participants with significant experience in interacting with individuals diagnosed with ASD in various real-life settings, including therapists, teachers, and parents of children with ASD. The research findings revealed a broad spectrum of technologies involved in ASD interventions, including applications, devices, and robots. The results bring a new perspective on the interactive technologies used in the therapy and diagnosis of ASD and highlight their important characteristics that can serve as a standard in the development of future technological solutions.

Author 1: Mihaela Chistol

Author 2: Mirela Danubianu

Author 3: Adina-Luminita Barîla

Keywords: Autism spectrum disorder; technology-mediated interventions; assistive technologies; therapy; cross-sectional study

Paper 6: Drier Bed Adsorption Predictive Model with Enhancement of Long Short-Term Memory and Particle Swarm Optimization

Abstract: The drier bed adsorption processes remove moisture from gases and liquids by ensuring product quality, extending equipment lifespan, and enhancing safety in various applications. The longevity of adsorption beds is quantified by net loading capacity values that directly impact the effectiveness of the moisture removal process. Predictive modeling has emerged as a valuable tool to enhance drier bed adsorption systems. Despite the increasing significance of predictive modeling in enhancing the efficiency of drier bed adsorption processes, the existing methodologies frequently exhibit deficiencies in accuracy and flexibility, which are crucial for optimizing process performance. This research investigates the effectiveness of a hybrid approach combining Long Short-Term Memory and Particle Swarm Optimization (LSTM+PSO) as a proposed method to predict the net loading capacity of a drier bed. The train-test split ratios and rolling origin technique are explored to assess model performance. The findings reveal that LSTM+PSO with a 70:30 train-test split ratio outperform other methods with the lowest error. Bed 1 exhibits an RMSE of 1.31 and an MSE of 0.91, while Bed 2 archives RMSE and MSE values of 0.81 and 0.72, respectively and Bed 3 with an RMSE of 0.19 and an MSE of 0.13, followed by Bed 4 with an RMSE of 0.67 and an MSE of 0.36. Bed 5 exhibits an RMSE of 0.42 and an MSE of 0.34. Furthermore, this research compares LSTM+PSO with LSTM and conventional predictive methods: Support Vector Regression, Seasonal Autoregressive Integrated Moving Average with Exogenous Variables, and Random Forest.

Author 1: Marina Yusoff

Author 2: Mohamad Taufik Mohd Sallehud-din

Author 3: Nooritawati Md. Tahir

Author 4: Wan Fairos Wan Yaacob

Author 5: Nur Niswah Naslina Azid

Author 6: Jasni Mohamad Zain

Author 7: Putri Azmira R Azmi

Author 8: Calvin Karunakumar

Keywords: Adsorption; Long Short-Term Memory; net loading capacity; Particle Swarm Optimization; prediction

Paper 7: A Deep Learning-based Approach for Vision-based Weeds Detection

Abstract: Weed detection is an essential component of smart agriculture, and the use of remote sensing technologies has the potential to significantly improve weed management practices, reduce herbicide usage, and increase crop yields. This study proposed an approach to weed detection using computer vision and deep learning technologies. By utilizing remote sensing methods based on DL, this approach has the potential to optimize weed management strategies, minimize herbicide use, and enhance crop productivity. The weed detection algorithm is based on the Yolov8 framework, and a custom model is trained using images from popular datasets as well as the internet. To evaluate the model's effectiveness, it is tested on both validation and testing sets. Furthermore, the model's performance is assessed using images that are not included in the original dataset. As experimental results shown, the deep learning-based approach is a promising solution for weed detection in agriculture.

Author 1: Yan Wang

Keywords: Smart agriculture; weed detection; remote sensing; deep learning; computer vision

Paper 8: Detection of Fruit using YOLOv8-based Single Stage Detectors

Abstract: In the agricultural sector, the precise detection of fruits plays a pivotal role in optimizing harvesting procedures, minimizing waste, and ensuring the delivery of high-quality produce. Deep learning methods have consistently exhibited superior accuracy compared to alternative techniques, making them a focal point in fruit detection research. However, the ongoing challenge lies in meeting the stringent accuracy requirements essential for real-world applications in agriculture. Addressing this critical concern, this study proposes an innovative solution utilizing the Yolov8 architecture for fruit detection. The methodology involves the meticulous creation of a custom dataset tailored to capture the diverse characteristics of agricultural fruits, followed by rigorous training, validation, and testing processes. Through extensive experimentation and performance evaluations, the findings underscore the exceptional accuracy achieved by the Yolov8-based model. This methodology not only surpasses existing benchmarks but also establishes a robust foundation for transforming fruit detection practices in agriculture. By effectively addressing the challenges associated with accuracy rates, this approach opens new avenues for optimized harvesting, waste reduction, and enhanced efficiency in agricultural practices, contributing significantly to the evolution of precision farming technologies.

Author 1: Xiuyan GAO

Author 2: Yanmin ZHANG

Keywords: Fruit detection; agricultural sector; deep learning; YOLOv8 model; precision agriculture

Paper 9: Optimizing Mobile Ad Hoc Network Routing using Biomimicry Buzz and a Hybrid Forest Boost Regression - ANNs

Abstract: A mobile ad hoc network (MANET) is a network of moving nodes that can interact with one another without the aid of a centrally located infrastructure. In MANETs, every node acts as a router and as a host, generating and consuming data. However, due to the mobility of nodes and the absence of centralized control, the routing process in MANETs is challenging. Therefore, routing protocols in MANETs are required to be efficient, scalable, and adaptable to the dynamic topology changes of the network. This paper proposes an optimized route selection approach for MANETs via the biomimicry buzz algorithm with the Bellman-Ford-Dijkstra algorithm to improve the effectiveness and accuracy of the routing process. By integrating these behaviors into the algorithm, the approach can select the shortest path in a network, leading to an optimal routing solution. Furthermore, the paper explores the use of Forest Boost Regression (FR), a novel machine learning algorithm, to predict energy consumption in MANETs. Utilizing this will help the network run more efficiently and last longer. Additionally, the paper discusses the use of Artificial Neural Networks (ANNs) to forecast link failure in MANET s, thereby increasing network performance and dependability. The proposed work presents the experimental evaluation by using Ns-3 as the simulation tool. The experimental results indicate a variation in packet delivery ratio from 97% to 90%, an average end-to-end delay of approximately 19 ms, an increase in node speed energy consumption from 60 to 87 joules, and a simulation time energy consumption of 89 joules over 60 seconds. These results provide insights into the performance and efficiency of the proposed strategy in the context of MANETs.

Author 1: D Dhinakaran

Author 2: S. M. Udhaya Sankar

Author 3: S. Edwin Raja

Author 4: J. Jeno Jasmine

Keywords: MANET; routing protocols; optimized route selection; regression; machine learning; Artificial Neural Networks

Paper 10: Conceptualizing an Inductive Learning Situation in Online Learning Enabled by Software Engineering

Abstract: Our work highlights the importance of adopting a systematic and methodical software engineering approach to the development of information technology projects for e-learning. We place particular emphasis on conceptualizing pedagogical scenarios and an inductive online learning situation. To ensure effective management of the information systems development process, we applied instructional design principles and adopted the 2TUP process, a refined version of the Rational Unified Process (RUP) suitable for projects of all sizes. To provide a visual representation of the system architecture and inform instructional design decisions, we used the Unified Modeling Language (UML) to create class, use case, activity, and sequence diagrams. We aim to demonstrate the potential of a structured software engineering approach to creating effective and efficient e-learning systems by conceptualizing an inductive online learning situation and five concrete examples illustrating the system's functionality. Our work underlines the importance of using standardized modeling languages such as UML to facilitate communication between stakeholders and collaboration between instructional designers and software developers.

Author 1: Ouariach Soufiane

Author 2: Khaldi Maha

Author 3: Khaldi Mohamed

Keywords: Software engineering approach; instructional design; conceptualization scenario; online learning situation; inductive approach

Paper 11: The Role of AI in Mitigating Climate Change: Predictive Modelling for Renewable Energy Deployment

Abstract: This study looks at how AI algorithms like Random Forest, Support Vector Machines (SVM), and Deep Boltzmann Machine (DBM) can be used for predictive modeling to make it easier to use renewable energy sources while reducing the negative effects of climate change. Predictive models based on Artificial Intelligence show possible ways to get the most out of green energy sources, which could lead to fewer carbon emissions. The results of the preliminary studies show that these AI systems can make accurate predictions about how green energy will be made because they are good at making predictions and generalizing. This feature makes it possible to use resources effectively, which improves the reliability of the grid and encourages more people to use green energy sources. Ultimately, employing these AI programs will serve as powerful tools in combating climate change and fostering a more sustainable and eco-friendly environment.

Author 1: Nawaf Alharbe

Author 2: Reyadh Alluhaibi

Keywords: Renewable energy; climate change; predictive models; and Artificial Intelligence (AI)

Paper 12: An Exploratory Analysis of using Chatbots in Academia

Abstract: With the advancement of technology in this era, chatbots have become more than just robots, as they used to conduct time-consuming and labor-intensive routine tasks. Now, it is more than just a robot for routine duties; it interacts and produces like a human. Despite the efficacy and productivity of chatbots like ChatGPT-4 and Bard, there will be significant ethical implications for the academic community, particularly students and researchers. The current study is experimenting with ChatGPT-4 and Bard by producing scientific articles with specific criteria, then applying topic modeling to assess the extent to which the content of the articles is related to the required topic, and verifying references, plagiarism, and the accuracy of the chatbot-generated articles. The results indicated that the content is relevant to the topic, and the accuracy of ChatGPT-4 is greater than Bard. ChatGPT-4 achieved 96%, and the majority of the bibliographies are accurate, whereas Bard achieved 52%, and the majority of bibliographies are incorrect, and some are not available. It is unethical to rely on a chatbot to produce scientific content, despite its accuracy, because it is not as accurate as humans and requires a thorough review of the content it generates. Furthermore, it alters his responses based on the individual he is interrogating, regardless of whether his answers are correct, as he is unable to defend his knowledge.

Author 1: Njood K Al-harbi

Author 2: Amal A. Al-shargabi

Keywords: AI; chatbots; ChatGPT; GPT-4; bard; ethics; machine learning; topic modeling

Paper 13: A Comparative Study of Stemming Techniques on the Malay Text

Abstract: Text stemming, an essential preprocessing step in the development of Natural Language Processing (NLP) applications, involves the transformation of various word forms into their root words. Stemming plays a critical role in decreasing the volume of text, thereby enhancing the efficiency of various computational tasks such as information retrieval, text classification, and text clustering. Stemming is a rule-based approach. On the other hand, it frequently suffers affixation errors that result in under-stemming, over-stemming, or both, as well as unstemmed or spelling exceptions. Every language has different stemming techniques, and among the most well-known Malay stemming algorithms are the Othman and Ahmad algorithms. Therefore, this study aims to compare the performance of the stemming errors between the Othman and Ahmad algorithms in stemming Malay text, particularly on two different domains of textual datasets, which are the course summaries of the education domain and housebreaking crime reports of the crime domain. The Othman algorithm presents a set of 121 stemming rules (set A). In the meantime, Ahmad's algorithm proposes two distinct sets of stemming rules, comprising 432 (set B) and 561 rules (set C), respectively. Based on the experiment results with 100 course summaries, the Ahmad algorithm (Set B) obtained a higher accuracy rate of 93.61%. The second highest is the Ahmad algorithm (Set C) with 93.53%. The Othman algorithm achieved the lowest accuracy with 86.04% compared to the other two algorithms. Meanwhile, findings from the experiment with 100 housebreaking crime reports show similar results, with the Ahmad algorithm (Set C) achieving the highest stemming accuracy of approximately 93.80% and the Othman algorithm producing the lowest stemming accuracy (83.09%). The result indicates that stemming accuracy is consistent across different types of datasets.

Author 1: Rosmayati Mohemad

Author 2: Nazratul Naziah Mohd Muhait

Author 3: Noor Maizura Mohamad Noor

Author 4: Nur Fadilla Akma Mamat

Keywords: Algorithm; ahmad algorithm; malay language; othman algorithm; rule-based; stemming; stemmer

Paper 14: Advanced Detection of COVID-19 Through X-ray Imaging using CovidFusionNet with Hybrid CNN Fusion and Multi-resolution Analysis

Abstract: The rapid diagnosis of COVID-19 through imaging is crucial in the current pandemic scenario. This study introduces the CovidFusionNet, a novel model adapted for efficient COVID-19 image classification. By effectively combining fusing features from seven pre-trained convolutional neural networks (CNNs), our model presents better accuracy in detecting COVID-19 from X-ray images. Three separate datasets, obtained from Kaggle, were used in this study to ensure the reliability and robustness of the model. The Continuous and Discrete Wavelet Transform was implemented for robust multi-resolution image analysis to maintain image properties after denoising. A novel enhancement method was also proposed, combining the capabilities of Adaptive Histogram Equalization (AHE) and Wavelet Transforms to emphasize finer details and concurrently heighten clarity while minimizing noise. Furthermore, to mitigate class imbalance, an oversampling approach was implemented. Comprehensive validation using 12 metrics across each dataset verified the proposed consistent performance, with remarkable accuracies of 98.02% for Dataset One, 99.30% for Dataset Two, and 98.25% for Dataset Three. Comparing CovidFusionNet against seven well-known pre-trained models showed that CovidFusionNet appeared more capable. This research advances the area of image-based diagnosis using COVID-19 and provides a model for quick medical actions.

Author 1: Majdi Khalid

Keywords: COVID-19 diagnosis; X-ray imaging; wavelet transform; Adaptive Histogram Equalization (AHE); oversampling; image denoising; image classification

Paper 15: Implementation of a Convolutional Neural Network (CNN)-based Object Detection Approach for Smart Surveillance Applications

Abstract: In the realm of smart surveillance systems, a fundamental technique for tracking and evaluating consumer behavior is object detection through video surveillance. While existing research underscores object detection through deep learning techniques, a notable gap exists in adapting these methods to effectively capture and recognize small, intricate objects. This study addresses this gap by introducing a customized methodology tailored to meet the nuanced requirements of accurate and lightweight detection for small objects, especially in scenarios prone to visual complexity and object similarity challenges. The primary objective is to furnish a vision-based object identification method designed for surveillance applications in smart stores, with a particular focus on locating jewelry objects. To achieve this, a Convolutional Neural Network (CNN)-based object detector utilizing YOLOv7 is employed for precise object detection and location extraction. The YOLOv7 network undergoes rigorous training and verification on a unique dataset specifically curated for this purpose. Experimental results affirm the efficacy of the proposed object identification method, demonstrating its capacity to detect items relevant to smart surveillance applications.

Author 1: Weiguo Ni

Keywords: Smart surveillance; lightweight object detection; YOLOv7; small object recognition; vision-based identification

Paper 16: An Energy Efficient Routing Algorithm using Chaotic Grey Wolf with Mobile Sink-based Path Optimization for Wireless Sensor Networks

Abstract: The task of deploying an energy-conscious wireless sensor networks (WSNs) is challenging. One of the most effective methods for conserving WSNs energy is clustering. The deployed sensors are divided into groups by the clustering algorithm, and each group's cluster head (CH) is chosen to gather and combine data from other sensors in the group. Mobile Wireless Sensor Networks, which enable moving the sink node, aid in reducing energy consumption. Thus, this paper introduces an energy efficient clustering algorithm and optimized path for a mobile sink using a swarm intelligence algorithms. The Chaotic Grey Wolf Optimization (CGWO) approach is used to form clusters and identify CHs. While utilizing the Slime Mould Algorithm (SMA) for determining the shortest path between a mobile sink and CHs. The effectiveness of the suggested routing strategy is evaluated against that of other current, cutting-edge protocols. The findings demonstrate that in terms of overall energy consumption and network lifetime, the suggested algorithm performs better than others. While for stability period the proposed algorithm outperforms three of compared algorithms and was close to the fourth.

Author 1: Latifah Alharthi

Author 2: Alaa E. S. Ahmed

Author 3: Mostafa E. A. Ibrahim

Keywords: Wireless sensor network; clustering algorithm; grey wolf optimizer; slime mould algorithm; mobile sink

Paper 17: Research on Qubit Mapping Technique Based on Batch SWAP Optimization

Abstract: The conventional approach for initial qubit mapping in the Noisy Intermediate-Scale Quantum (NISQ) era typically uses a static heuristic strategy, overlooking insufficient qubit neighborhood in subsequent operations, resulting in excess additional SWAP gates. To address this, we introduce a multifactor interaction cost function considering qubit distance, interaction time, and gate operation error rates, enhancing SWAP gate selection in the traditional strategy. Considering quantum hardware constraints, we propose Batch SWAP Optimization Strategy (BSOS). BSOS tackles qubit mapping challenges by leveraging optimal SWAP gate selection and a SWAP-based batch update technique, effectively minimizing SWAP gates throughout circuit execution. Experimental results show that BSOS significantly reduces additional gates by intelligently selecting SWAP gates and using batch updating, with a 38.1% average decrease in inserted SWAP gates, leading to a 12% reduction in hardware gate counting overhead.

Author 1: Hui Li

Author 2: Kai Lu

Author 3: Ziao Han

Author 4: Huiping Qin

Author 5: Mingmei Ju

Author 6: Shujuan Liu

Keywords: Quantum computing; quantum circuit compilation; initial qubit mapping; Batch SWAP Optimization Strategy (BSOS); best SWAP choice; Batch Update Technology (BUT)

Paper 18: Identifying Factors in Congenital Heart Disease Transition using Fuzzy DEMATEL

Abstract: The transition from pediatric to adult cardiology care is a pivotal moment in the healthcare journey of individuals with congenital heart conditions or childhood-onset heart diseases. This multifaceted process requires meticulous consideration of clinical, psychosocial, and logistical factors. This research aims to explore the critical criteria for transitioning pediatric patients to adult cardiology, delving into the challenges and opportunities inherent in this healthcare shift. The identified factors for successful transition, including age and developmental stage, medical complexity, cardiac function, psychosocial factors, insurance, and financial considerations, play integral roles in the transition process. Leveraging analytical methodologies, particularly the Fuzzy Decision-Making Trial and Evaluation Laboratory (DEMATEL), this study involves three experts who assess criteria linguistically, converted to Triangular Fuzzy Numbers, and averaged. Defuzzification, using the CFCS method, yields crisp values. Results reveal that Medical Complexity (U+V = 3.96, U-V = 0.233), Insurance (U+V = 3.931, U-V = 0.22), Psychosocial Factors (U+V = 3.839, U-V = 0.387), and Age and Developmental Stage (U+V = 3.802, U-V = 0.106) follow Cardiac Function (U+V = 4.312, U-V = 0.946) in ranking. Age and Developmental Stage, Medical Complexity, Psychosocial Factors, and Insurance are considered causal variables, with Cardiac Function as an effect. These numerical insights enhance our understanding of transition criteria interdependencies, informing tailored healthcare strategies.

Author 1: Raghavendra M Devadas

Author 2: Vani Hiremani

Author 3: Ranjeet Vasant Bidwe

Author 4: Bhushan Zope

Author 5: Veena Jadhav

Author 6: Rohini Jadhav

Keywords: DEMATEL; Fuzzy DEMATEL; factors; pediatric patients; heart disease

Paper 19: Reliable and Efficient Model for Water Quality Prediction and Forecasting

Abstract: Water quality is a crucial aspect of environmental and public health. Hence, its assessment is of paramount importance. This research paper aims to leverage machine learning models to classify water quality based on a comprehensive dataset. The dataset contains various water quality indicators, and the primary objective is to predict whether the water is safe or not to consume or use. This research evaluates the performance of diverse machine learning algorithms, such as Decision Trees, Random Forest, Logistic Regression, Support Vector Machines, and more for comparative analysis. Performance metrics such as accuracy, precision, recall, and F1-score are used to assess the models' effectiveness in classifying water quality. The Random Forest algorithm gave the best performance with an accuracy of 95.08%, an F1-Score of 94.69%, a Precision of 90.48%, a Recall of 93.10%, and an AUC score of 0.91. A comparative plot for the ROC AUC curve is also plotted between the various machine learning models used. Feature importance, which can help identify which water quality parameters have the greatest impact on predicting water quality outcomes, is also found in the research work.

Author 1: Azween Abdullah

Author 2: Himakshi Chaturvedi

Author 3: Siddhesh Fuladi

Author 4: Nandhika Jhansi Ravuri

Author 5: Deepa Natesan

Author 6: M. K Nallakaruppan

Keywords: Random forest; logistic regression; feature importance; decision trees; support vector machines

Paper 20: Data Mining Application Forecast of Business Trends of Electronic Products

Abstract: Sales forecasting is a pressing concern for companies amid rising consumer demand and intensifying competition, compounded by declining sales due to growing socio-economic challenges. Currently, many companies are having difficulty selling products due to a lack of management systems. To assist that, data mining techniques are introduced but it is difficult to evaluate the data and it is practically impossible to accurately forecast large amounts of data. However, data mining remains an important management tool that supports early decisions to increase profits, innovate business trends and improve sales by generating intelligence from the company's data resources. In this article, the research object chosen is the data of a nationwide electronics company. Their sales volume data for consumer electronics was used and applied to this study. The study used a "clustering" algorithm to group data based on the unique characteristics of each product, region, season, and time to estimate the amount of goods sold in the past, thereby predicting the amount of goods that will be exported. Password is sold in the following years and look for market trends. For each group, the results obtained with k = 3 show that the number of elements in each cluster is 771422, 11874, and 312, respectively. Combined with the “regression tree” algorithm for cluster partitioning and using the protocol Evaluate MSE and RMSE to evaluate the accuracy of the model, a result of 43065.66 Sales forecasting results show that the model's accuracy is close to realistic accuracy and depends on seasonal factors that are really important to some people. Based on the above results, the business's marketing campaigns and strategies will be deployed and achieve high results.

Author 1: Kheo Chau Mui

Author 2: Nhon Nguyen Thien

Keywords: Data mining; sales forecasting; clusters; regression tree; RMSE; MSE; k-prototypes

Paper 21: Identification of Microaneurysms and Exudates for Early Detection of Diabetic Retinopathy

Abstract: Diabetic retinopathy (DR) is a condition that may be a complication of diabetes, and it can damage both the retina and other small blood vessels throughout the body. Microaneurysms (MA’s) and Hard exudates (HE’s) are two symptoms that occur in the early stage of DR. Accurate and reliable detection of MA’s and HE’s in color fundus images has great importance for DR screening. Here, a machine learning algorithm has been presented in this paper that detects MA’s and HE’s in fundus images of the retina. In this research a dynamic thresholding and fuzzy c mean clustering with characteristic feature extraction and different classification techniques are used for detection of MA’s and HE’s. The performance of system is evaluated by computing the parameters like sensitivity, specificity, accuracy, and precision. The results are compared between different types of classifiers. The Logistic Regression classifier (LRC) performance is good when compared with other classifiers with an accuracy of 94.6% in detection of MA’s and 96.2% in detection of HE’s.

Author 1: G Indira Devi

Author 2: D. Madhavi

Keywords: Diabetic retinopathy; microaneurysms; hard exudates; SVM; LRC

Paper 22: Gamers Intention Towards Purchasing Game Items in Virtual Community: Extending the Theory of Planned Behavior

Abstract: Virtual communities serve as bustling marketplaces where gamers engage in transactions for in-game items, driving the digital economy's expansion. This research aims to illuminate the determinants of steering users' decisions within these online environments. Focusing on the constructs of Attitude, Subjective Norms, and Perceived Behavioral Control derived from the Theory of Planned Behavior (TPB), we investigate the factors shaping purchasing intentions. Employing structural equation modeling (SEM) on a robust dataset of 300 validated respondents, our analysis unveils insights into user motivations. Notably, the amalgamation of Attitude, Subjective Norms, and Perceived Behavioral Control explains 84% of the drivers guiding in-game item transactions within virtual communities. Our findings underscore the significance of certain attributes. Specifically, the perceived wisdom inherent in these transactions, the constructive influence of community discussions, and the ease of communication and negotiation channels within virtual realms emerge as pivotal determinants influencing user behavior. This study not only contributes to understanding user behavior in virtual spaces but also holds practical implications for scholars and industry stakeholders. By shedding light on these influential factors, this research informs strategies and interactions within virtual communities, offering valuable insights into the dynamics of the digital marketplace.

Author 1: Abbi Nizar Muhammad

Author 2: Achmad Nizar Hidayanto

Keywords: Theory of planned behavior; in-game items; purchasing intention; virtual community

Paper 23: Deep Learning-based License Plate Recognition in IoT Smart Parking Systems using YOLOv6 Algorithm

Abstract: License plate recognition (LPR) is pivotal for the seamless operation of Internet of things (IoT) and smart parking systems, ensuring the swift and effective identification and management of vehicles. Recent research has concentrated on refining LPR methods through deep learning approaches, proposing diverse strategies to enhance accuracy and reduce computation costs. This work tackles these challenges by introducing an innovative method rooted in the YOLOv6 algorithm. Leveraging a tailored dataset for model generation, the study employs rigorous methodologies involving validation, testing, and training. The resultant model demonstrates marked improvements in license plate recognition capabilities, surpassing the performance of existing methods. This breakthrough bears significant implications for advancing IoT smart parking systems, promising heightened reliability and efficiency in vehicle identification and management. Thorough experimental results and performance evaluations validate the efficacy of the proposed YOLOv6-based method. In-depth discussions and comparisons with state-of-the-art methods in the field lead to the conclusion that the introduced approach not only elevates accuracy but also enhances overall efficiency in license plate recognition for smart parking systems, thereby providing valuable contributions to the domain.

Author 1: Ming Li

Author 2: Li Zhang

Keywords: Internet of things; deep learning; smart parking system; license plate recognition; YOLOv6

Paper 24: Influence of Membership Function and Degree on Sorghum Growth Prediction Models in Machine Learning

Abstract: Rapid advances in science and technology have significantly changed plant growth modeling. The main contribution to this transformation lies in using Machine Learning (ML) techniques. This study focuses on sorghum, an important agricultural crop with significant economic implications. Crop yield studies include temperature, humidity, climate, rainfall, and soil nutrition. This research has a novelty: the input factors for predicting sorghum plant growth, namely the treatment of applying organic fertilizer and dolomite lime to sorghum planting land. The three predicted sorghum plant growth factors, namely Height, Biomass, and Panicle weight, are the reasons for using the Multiple Adaptive Neural Fuzzy Inference System (MANFIS) model. This research investigates the impact of Membership Function and Degree on the MANFIS model. A comprehensive comparison of various membership functions, including Gaussian, Triangular, Bell, and Trapezoidal functions, along with various degrees of membership, has been carried out. The dataset used includes data related to sorghum growth obtained from field experiments. The main objective was to assess the effectiveness of membership and degree functions in accurately predicting sorghum growth parameters, consisting of height, biomass, and panicle weight. This assessment uses metrics such as MAPE (Mean Absolute Percentage Error), MAE (Mean Absolute Error), and RMSE (Root Mean Square Error) to evaluate the predictive performance of the MANFIS model when using four different types of membership functions and degrees. The results obtained the best level of accuracy in predicting panicle weight (ANFIS-3) with chicken manure treatment using the Trapezoidal membership function type and degree of membership function [3,3] with MAPE results of 5.77%, MAE of 0.2994, and RMSE of 0.395.

Author 1: Abdul Rahman

Author 2: Ermatita

Author 3: Dedik Budianta

Author 4: Abdiansah

Keywords: Prediction; MANFIS; membership function; organic fertilizer; sorghum

Paper 25: A Hybrid Double Encryption Approach for Enhanced Cloud Data Security in Post-Quantum Cryptography

Abstract: Quantum computers and research on quantum computers are increasing due to the efficiency and speed required for critical applications. This scenario also kindles the vitality of data protection needed against threats from quantum computers. Research in post-quantum threats is very minimal so far, but it is much needed to protect the enormous data stored in the cloud for healthcare, governmental, or any crucial data. This research work presents an advanced hybrid double encryption approach for cloud data security based on Post-Quantum Cryptography (PQC) to ensure the restriction of unauthorized access. The suggested approach combines the benefits of the NTRU encryption and AES encryption algorithms and works in hybrid mode, offering strong security while resolving issues with real-time performance and cost-efficiency. A streamlined key management system is set together to improve real-time processing, significantly reducing encryption and decryption delay times. Moreover, NTRU Encrypt dynamic parameter selection, which adapts security parameters based on data sensitivity, maintains accurate information and security. In addition to addressing real-time performance and data security, an innovative development in this method is known as Quantum-Adaptive Stream Flow Encryption (QASFE), which enables secure data sharing and collaborative working within a quantum-resistant framework. This innovative feature enhances data accessibility while maintaining the highest level of security. In the era of post-quantum cryptography, our multifactor authentication technique, integrating double encryption and QASFE, is a proactive and flexible solution for securing cloud data, and protecting data security and privacy against emerging threats.

Author 1: Manjushree C V

Author 2: Nandakumar A N

Keywords: Cloud data security; double encryption; Post-Quantum Cryptography (PQC); NTRU Encrypt; AES Encryption

Paper 26: Comprehensive Analysis of Topic Models for Short and Long Text Data

Abstract: The digital age has brought significant information to the Internet through long text articles, webpages, and short text messages on social media platforms. As the information sources continue to grow, Machine Learning and Natural Language Processing techniques, including topic modeling, are employed to analyze and demystify this data. The performance of topic modeling algorithms varies significantly depending on the text data's characteristics, such as text length. This comprehensive analysis aims to compare the performance of the state-of-the-art topic models: Nonnegative Matrix Factorization (NMF), Latent Dirichlet Allocation using Variational Bayes modeling (LDA-VB), and Latent Dirichlet Allocation using Collapsed Gibbs-Sampling (LDA-CGS), over short and long text datasets. This work utilizes four datasets: Conceptual Captions and Wider Captions, image captions for short text data, and 20 Newsgroups news articles and Web of Science containing science articles for long text data. The topic models are evaluated for each dataset using internal and external evaluation metrics and are compared against a known value of topic 'K.' The internal and external evaluation metrics are the statistical metrics that assess the model's performance on classification, significance, coherence, diversity, similarity, and clustering aspects. Through comprehensive analysis and rigorous evaluation, this work illustrates the impact of text length on the choice of topic model and suggests a topic model that works for varied text length data. The experiment shows that LDA-CGS performed better than other topic models over the internal and external evaluation metrics for short and long text data.

Author 1: Astha Goyal

Author 2: Indu Kashyap

Keywords: Topic modeling; Nonnegative Matrix Factorization (NMF); Latent Dirichlet Allocation (LDA); evaluation metrics; short text mining; long text mining

Paper 27: Assessing User Requirements for e-Resources Interface Design in University Libraries

Abstract: e-Resources in the university library as learning resources are one of the primary services that promote learning and research to improve university productivity. At present, users find it difficult to access e-resources and require assistance in finding them. When using the system, users felt frustrated, confused, and lost. The e-resources services system on library websites, on the other hand, lacks sociability and a sense of human warmth. Sociability and a sense of human warmth can be integrated into the website interface, which may evoke the sensation of being with an actual individual, even if the service is provided online. This study investigated the social presence aspects that can be implemented in the library's e-resources system. The purpose of this study is to elicit social presence features that can be implemented in the design of e-resource interfaces on library websites. The methods used in this study are in three phases: a) web content analysis from twelve university library interfaces designed in several countries; b) interviews with library staff; and c) assessment by a questionnaire of library website users. Website content analysis was used to investigate elements that offer many unique features to support the implementation of social presence through the e-resources interface. An interview was used to validate elements that were found in the web content analysis, and a questionnaire phase was used to assess the user requirements for these social presence elements. The results of empirical studies show that users need some elements of social presence, such as comments, chat, ratings, voice, personalized welcome in library accounts, tools, preference language, links for reference managers, and social media, as well as ease of access such as readable help font, color, and font size.

Author 1: Yuli Rohmiyati

Author 2: Tengku Siti Meriam Tengku Wook

Author 3: Noraidah Sahari

Author 4: Siti Aishah Hanawi

Keywords: User requirement; interface design; e-resources; social presence; university library; element

Paper 28: Multidimensional Private Information Portrait in Social Network Users

Abstract: In order to tackle the challenges of users' weak privacy awareness and frequent disclosure of private information in social network, this paper proposes a multidimensional privacy information portrait model of users in Chinese social networks. Because the TF-IDF (Term Frequency-Inverse Document Frequency, TF-IDF) algorithm does not consider the distribution of feature terms among and within classes, uses the TF-IDF algorithm based on the bag-of-words model to calculate the sensitivity of user privacy information. Considering the diversity of user privacy information, this paper proposes the PROLM (Positive reverse order lookaround matching ) algorithm, which is combined with the Flashtext+ (improved Flashtext) algorithm and SMA (string matching algorithm, SMA), the PROLM_FlashText+_SMA to extract user personal privacy information and location where the privacy information is located, and return the sensitivity. Using the BERT (Bidirectional Encoder Representation from Transformers, BERT)-Softmax privacy information classification model, the privacy information is classified into high, moderate and mild privacy information, and a multidimensional privacy information portrait of the user is constructed based on the privacy information and sensitivity. The experiments show that the accuracy of PROLM_FlashText+_SMA algorithm for privacy information extraction reaches 93.63%, and the overall F1 index of privacy information classification using the BERT-Softmax model reaches 0.9798 on the test set, better than baseline comparison model, has better privacy information classification effect.

Author 1: Fangfang Shan

Author 2: Mengyi Wang

Author 3: Huifang Sun

Keywords: Social network; personal privacy information; privacy information portrait; sensitivity; privacy protection; BERT

Paper 29: Smart Fruit Identification and Counting using Machine Vision Approach

Abstract: Estimating fruit yield holds significant importance for farmers as it enables them to make precise resource management decisions for fruit harvesting. The adoption of automated image processing technology not only reduces the human labor required but also enhances the accuracy of ripe fruit estimates. This research delves into the performance of an image processing algorithm designed to count and identify oranges. The study employed a multi-phase approach, starting with the creation of a mask to isolate orange content, followed by the detection of circular shapes within the mask. Lastly, the algorithm filters and counts the identified circles. The outcome of this study revealed that the algorithm demonstrated an impressive success rate of approximately 72.4% in correctly identifying oranges with standard deviation of +/- 12.20.

Author 1: Madhura R. Shankarpure

Author 2: Dipti D. Patil

Keywords: Image processing; fruit; multiphase approach; counting

Paper 30: Towards a Framework for Elevating the Usage of eLearning Technologies in Higher Education Institutions

Abstract: Adopting eLearning technologies is no longer an option in Higher Education Institutions (HEIs) to support teaching and learning activities. However, despite steps taken by most of these institutions towards effective utilization of technologies, the current situation on the ground shows two significant challenges. First, the misalignment between institutions’ strategies and the technical implementation of these technologies. Second, the scattered implementation and usage of eLearning technologies among stakeholders, which increases operational overhead due to the lack of a unified approach and usage procedures that promote optimal utilization of such technologies. This paper aims to introduce a framework for elevating the usage of eLearning technologies in HEIs. It guides the alignment between strategic goals and technology implementation for effective and progressive eLearning technology usage. Design science research methodology is adopted to guide the development of this framework. It drives the development process by first being aware of the problem from a real-life context and then proposing a solution. Principles from business and IT alignment and enterprise architecture are adopted to propose this framework, which is meant to be comprehensive to have eLearning technologies fit the institution’s purpose while achieving strategic goals.

Author 1: Naif Alzahrani

Author 2: Hassan Alghamdi

Keywords: eLearning; higher education; business and IT alignment; enterprise architecture

Paper 31: A Novel Paradigm for IoT Security: ResNet-GRU Model Revolutionizes Botnet Attack Detection

Abstract: The rapid proliferation of the Internet of Things (IoT) has engendered substantial security apprehensions, chiefly due to the emergence of botnet attacks. This research study delves into the realm of Intrusion Detection Systems (IDS) by leveraging the IoT23 dataset, with a specific emphasis on the intricate domain of IoT at the network's edge. The evolution of edge computing underscores the exigency for tailored security solutions. An array of statistical methodologies, encompassing ANOVA, Kruskal-Wallis, and Friedman tests, is systematically employed to illuminate the evolving trends across multiple facets of the study. Given the intricacies entailed in feature selection within edge environments, Chi-square analyses, Recursive Feature Elimination (RFE), and Lasso-based techniques are strategically harnessed to unearth meaningful feature subsets. A meticulous evaluation encompassing 19 classifiers, meticulously selected from both machine learning (ML) and deep learning (DL) paradigms, is rigorously conducted. Initial findings underscore the potential of the Gated Recurrent Unit (GRU) model, especially when coupled with intrinsic lasso-based feature selection. This promising outcome catalyzes the formulation of an ensemble approach that harnesses multiple LassoCV models, aimed at amplifying feature selection proficiency. Furthermore, an optimized ResNet-GRU model emerges from the fusion of the GRU and ResNet architectures, with the objective of augmenting classification performance. In response to mounting concerns regarding data privacy at the edge, a resilient federated learning ecosystem is meticulously crafted. The seamless integration of the optimized ResNet-GRU model into this framework facilitates the employment of FedAvg, a widely acclaimed federated learning methodology, to adeptly navigate the intricacies associated with data sharing challenges. A comprehensive performance evaluation is undertaken, wherein the ResNet-GRU model is benchmarked against FedAvg and a diverse array of other federated learning algorithms, including FedProx and Fed+. This extensive comparative analysis encompasses a spectrum of performance metrics and processing time benchmarks, shedding comprehensive light on the capabilities of the model.

Author 1: Jyotsna A

Author 2: Mary Anita E. A

Keywords: Internet of things; federated learning; Gated Recurrent Neural Networks; Long Short Term Memory (LSTM)

Paper 32: Hybrid Approach with VADER and Multinomial Logistic Regression for Multiclass Sentiment Analysis in Online Customer Review

Abstract: Sentiment analysis is crucial for businesses to understand customer reviews and assess sentiment polarity. A hybrid technique combining VADER and Multinomial Logistic Regression was used to analyze customer sentiment in online customer review data. VADER is a lexicon-based approach that labels reviews with sentiment using a predefined lexicon, whereas Multinomial Logistic Regression can determine the polarity of sentiment using VADER data. This study employed multiclass classification using TF-IDF vectorization to categorize sentiment as a positive, negative, or neutral class. Correctly managing neutral sentiments can assist businesses in identifying improvement opportunities. The utilization of the VADER lexicon and Multinomial Logistic Regression has been shown to significantly improve the performance of sentiment analysis in the context of multiclass classification problems. With a 75.213% accuracy rate, the VADER lexicon accurately recognizes neutral sentiment and is appropriate to adapt in categorizing sentiment related to customer reviews. Combined with Multinomial Logistic Regression, accuracy increases to 92.778%. In conclusion, the hybrid approach with VADER and Multinomial Logistic Regression can leverage the accuracy and reliability of multiclass customer sentiment analysis.

Author 1: Murahartawaty Arief

Author 2: Noor Azah Samsudin

Keywords: Hybrid approach; multiclass sentiment analysis; VADER; multinomial logistic regression; online customer review

Paper 33: Artificial Intelligence-based Optimization Models for the Technical Workforce Allocation and Routing Problem Considering Productivity

Abstract: Ensuring the reliability and availability of electric power networks is essential due to the increasing demands. An effective preventive maintenance strategy requires efficient resources allocation to perform the maintenance tasks, particularly the technical workforce. This paper introduces an innovative artificial intelligence-based approach to predict workforce productivity, aiming to optimize both the allocation of the technical workforce for maintenance tasks and their routing. In this study, two mathematical optimization models are introduced that utilize the output value of Artificial Neural Networks (ANN) for optimal resource allocation and routing. The first model focuses on team formation, considering the predicted productivity in order to ensure effective collaboration. While the second model focuses on the optimal assignment and routing of these teams to specific maintenance tasks. Validated with real-world data, the models show considerable promise in enhancing resource allocation, task assignment, and cost-efficiency in the electricity industry. Furthermore, sensitivity analysis has been conducted and managerial insights has been explored. The study also paves the way for future research, highlighting the potential for refining these models for more extensive applications.

Author 1: Mariam Alzeraif

Author 2: Ali Cheaitou

Keywords: Productivity; workforce; maintenance; optimization; allocation; routing

Paper 34: Development of Crack Detection and Crack Length Calculation Method using Image Processing

Abstract: To evaluate the integrity of a building, many experts and engineers have evaluated the damage classification of a building based on superficial visual information through field surveys. On-site surveys are hazardous and require several years of experience and expertise. In this study, a system for detecting the presence or absence of cracks and calculating their lengths was developed using image processing technology. The accuracy of the system was examined using crack image data obtained from shear force experiments. For crack detection, a crack detection method was developed using canny edge, threshold, and HSV color detection. The detection of the presence of cracks was proposed to be coupled with image segmentation to improve detection accuracy. A method for calculating the crack length using image processing was also developed. In this study, we proposed a method to calculate cracks as straight lengths, and obtained results with 98.1% accuracy. However, for curved cracks, it was necessary to rotate or segment the image.

Author 1: Jewon Oh

Author 2: Yutaka Matsumoto

Author 3: Kohei Arai

Keywords: Image processing; crack detection; length calculation; color detection; canny; threshold; OpenCV

Paper 35: Urban Image Segmentation in Media Integration Era Based on Improved Sparse Matrix Generation of Digital Image Processing

Abstract: Media integration integrates the resources of various media platforms, including audience, technical and human resources. In the era of media integration, media in various channels, at different levels and in different fields provide various choices for image communication and brand building of cities. The technology of image processing by computer has gradually affected all aspects of people’s life and work, bringing more and more convenience to people. In this paper, the application of digital image processing technology in city image communication in the age of media integration is studied. A new sparse matrix creation method is proposed, and the created sparse matrix is used as the similarity matrix to segment the spectral clustering image, so that the edge contour weakened in gradient calculation can be corrected and strengthened again. The research shows that the improved algorithm is superior to the traditional algorithm, and compared with the fuzzy entropy algorithm based on exhaustive search, the gray contrast between regions and Bezdek partition coefficient are improved by 9.301% and 4.127%. In terms of speed, the algorithm in this paper has absolute advantages, so our research is also affirmed, which fully shows that it should have high application value.

Author 1: Dan Zheng

Author 2: Yan Xie

Keywords: Media integration; digital image processing; city image; image segmentation; improved sparse matrix generation

Paper 36: Towards a Stacking Ensemble Model for Predicting Diabetes Mellitus using Combination of Machine Learning Techniques

Abstract: Diabetes Mellitus (DM) is a chronic disease affecting the world's population, it causes long-term issues such as kidney failure, blindness, and heart disease, hurting one's quality of life. Diagnosing diabetes mellitus in an early stage is a challenge and a decisive decision for medical experts, as delay in diagnosis leads to complications in controlling the progression of the disease. Therefore, this research aims to develop a novel stacking ensemble model to predict diabetes mellitus a combination of machine learning models, where an ensemble of Prediction classifiers was used, such as Random Forest (RF), Logistic Regression (LR), as base learners' models, and the Extreme gradient Boosting model (XGBoost) as a Meta-Learner model. The results indicated that our proposed stacking model can predict diabetes mellitus with 83% accuracy on Pima dataset and 97% with DPD dataset. In conclusion, our proposed model can be used to build a diagnostic application for diabetes mellitus, as recommend testing our model on a huge and diverse dataset to obtain more accurate results.

Author 1: Abdulaziz A Alzubaidi

Author 2: Sami M Halawani

Author 3: Mutasem Jarrah

Keywords: DM; Diabetes Mellitus; Stacking; Ensemble learning; Machine Learning; Random Forest (RF); Logistic Regression (LR); Extreme Gradient Boosting model (XGBoost)

Paper 37: Encryption Traffic Classification Method Based on ConvNeXt and Bilinear Attention Mechanism

Abstract: The rapid growth in internet traffic resulted to the emergence of network traffic categorization as a crucial area of research in network performance and management. This technological advancement has demonstrated its efficacy in aiding network administrators to identify anomalies within network behavior. However, the widespread adoption of encryption technology and the continual evolution of encryption protocols present a novel challenge in the classification of encrypted traffic. Addressing this challenge, this paper introduces an innovative methodology for classifying encrypted traffic by harnessing ConvNeXt and a fusion attention mechanism. Through the representation of traffic data as images and the integration of a bilinear attention mechanism into the model, our proposed approach attains heightened precision in the classification of encrypted network traffic. To substantiate the effectiveness of our methodology, experiments were conducted employing the publicly available ISCX VPN-nonVPN dataset. The experimental findings showcase superior recognition performance, underscoring the efficacy of the proposed approach.

Author 1: Xiaohua Feng

Author 2: Yuan Liu

Keywords: Encryption traffic recognition; end-to-end; convolutional neural network; bilinear attention module

Paper 38: Energy-Aware Clustering in the Internet of Things using Tabu Search and Ant Colony Optimization Algorithms

Abstract: The Internet of Things (IoT) significantly impacts communication systems' efficiency and the requirements for applications in our daily lives. Among the major challenges involved in data transmission over IoT networks is the development of an energy-efficient clustering mechanism. Recent methods are challenged by long transmission delays, imbalanced load distribution, and limited network lifespan. This paper suggests a new cluster-based routing method combining Tabu Search (TS) and Ant Colony Optimization (ACO) algorithms. The TS algorithm overcomes the disadvantage of ACO, in which ants move randomly throughout the colony in search of food sources. In the process of solving optimization problems, the ACO algorithm traps ants, resulting in a considerable increase in the time required for local searches. TS can be used to overcome these drawbacks. In fact, the TS algorithm eliminates the problem of getting stuck in local optima due to the randomness of the search process. Experimental results indicate that the proposed hybrid algorithm outperforms ACO, LEACH, and genetic algorithms regarding energy consumption and network lifetime.

Author 1: Mei Li

Author 2: Jing Ai

Keywords: Internet of things; clustering; data transmission; energy efficiency; ant colony optimization algorithm

Paper 39: Machine Learning-based Secure 5G Network Slicing: A Systematic Literature Review

Abstract: As the fifth-generation (5G) wireless networks continue to advance, the concept of network slicing has gained significant attention for enabling the provisioning of diverse services tailored to specific application requirements. However, the security concerns associated with network slicing pose significant challenges that demand comprehensive exploration and analysis. In this paper, we present a systematic literature review that critically examines the existing body of research on machine learning techniques for securing 5G network slicing. Through an extensive analysis of a wide range of scholarly articles selected from specific search databases, we identify and classify the key machine learning approaches proposed for enhancing the security of network slicing in the 5G environment. We investigate these techniques based on their effectiveness in addressing various security threats and vulnerabilities while considering factors such as accuracy, scalability, and efficiency. Our review reveals that machine learning techniques, including deep learning algorithms, have been proposed for anomaly detection, intrusion detection, and authentication in 5G network slicing. However, we observe that these techniques face challenges related to accuracy under dynamic and heterogeneous network conditions, scalability when dealing with a large number of network slices, and efficiency in terms of computational complexity and resource utilization. To overcome these challenges, our experimentation shows that the integration of reinforcement learning techniques with CNNs, multi-agent reinforcement learning, and distributed SVM frameworks emerged as potential solutions with improved accuracy and scalability in network slicing. Furthermore, we identify promising research directions, including the exploration of hybrid machine learning models, the adoption of explainable AI techniques, and the investigation of privacy-preserving mechanisms.

Author 1: Meshari Huwaytim Alanazi

Keywords: 5G; accuracy; deep learning; efficiency; security; machine learning; network slicing; scalability

Paper 40: Comparison of the Application of Weighted Cosine Similarity and Minkowski Distance Similarity Methods in Stroke Diagnostic Systems

Abstract: Stroke is a critical medical condition requiring prompt intervention due to its multifaceted symptoms and causes influenced by various factors, including psychological aspects and the patient's lifestyle or daily habits that impact risk factors. The recovery process involves consistent medical care and lifestyle adjustments tailored to the individual case. Expert Systems, a scientific field focused on studying and developing diagnostic systems, can employ the Case-based Reasoning method to identify the type of stroke based on similarities with prior patient cases, considering specific causes and symptoms. This study utilizes the Weighted Cosine, Jaccard Coefficient, and Minkowski Distance methods to assess the similarity of stroke cases. The evaluation is based on input data such as patient causes or symptoms and risk factors from medical records. The analysis of case similarity and solutions involves applying the Weighted Cosine, Jaccard Coefficient, and Minkowski Distance methods, with a defined threshold value. The highest similarity values from previous patient cases are selected for each method. The test outcomes suggest that employing the Minkowski Distance method with a threshold value of 75 and an r value of three or four yields the highest levels of accuracy, recall, and precision. The Minkowski Distance achieves an accuracy and recall rate of more than 88 percent with 100 percent precision.

Author 1: Joko Purwadi

Author 2: Rosa Delima

Author 3: Argo Wibowo

Author 4: Angelina Rumuy

Keywords: Expert system; stroke; case-based reasoning; Minkowski Distance; jaccard coefficient; weighted cosine; threshold; accuracy; diagnosis

Paper 41: Optimal Cluster Head Selection in Wireless Sensor Network via Combined Osprey-Chimp Optimization Algorithm: CIOO

Abstract: The development of Wireless Sensor Network (WSN) has gained significant attention for smart systems due to their potential use in a wide range of areas. WSN consists of tiny, independently arranged sensor nodes that run on batteries. The resources and energy usage for sensor nodes are the major factors. Particularly, the unbalanced nodes’ raises the energy use and reduces the network life-span. Energy efficiency in WSN cluster head selection remains a challenging task. The best method has been developed for reducing node energy consumption is clustering. However, the current clustering strategy failed to properly allocate the energy needs of the nodes without considering energy features, node quantity, as well as adaptability. Hence, there is need for advanced clustering process with new optimization tactics, and accordingly, a new cluster-head selection model in WSN is proposed in this work. Initially, the clustering process is done by the k-means algorithm. The Cluster Head (CH) selection is the subsequent progress under the consideration of node’s energy, distances, delays, and risks as well. A novel CIOO (Chimp Integrated Osprey Optimization) algorithm combining the Osprey and Chimp optimization algorithm is proposed for Cluster Head Selection (CHS). Finally, the performance of proposed model is evaluated over the conventional methods.

Author 1: Vikhyath K B

Author 2: Achyutha Prasad N

Keywords: Wireless sensor network; clustering; cluster head; cluster head selection; chimp optimization; osprey optimization

Paper 42: Deep Learning-based Pothole Detection for Intelligent Transportation: A YOLOv5 Approach

Abstract: Pothole detection plays a crucial role in intelligent transportation systems, ensuring road safety and efficient infrastructure management. Extensive research in the literature has explored various methods for pothole detection. Among these approaches, deep learning-based methods have emerged as highly accurate alternatives, surpassing other techniques. The widespread adoption of deep learning in pothole detection can be justified by its ability to learn discriminative features, leading to improved detection performance automatically. Nevertheless, the present research challenge lies in achieving high accuracy rates while maintaining non-destructiveness and real-time processing. In this study, we propose a deep learning model according to the YOLOv5 architecture to address this challenge. Our method includes generating a custom dataset and conducting training, validation, and testing processes. Experimental outcomes and performance evaluations show the suggested method's efficacy, showcasing its accurate detection capabilities.

Author 1: Qian Li

Author 2: Yanjuan Shi

Author 3: Qing Liu

Author 4: Gang Liu

Keywords: Pothole detection; deep learning; intelligent transportation systems; YOLOv5

Paper 43: Security and Privacy of Cloud Data Auditing Protocols: A Review, State-of-the-art, Open Issues, and Future Research Directions

Abstract: Cloud service providers offer a trustworthy and resistant-based storage environment for on-demand cloud services to outsource clients’ data. Several researchers and business entities currently adopt cloud services to store their data in remote cloud storage servers for cost-saving purposes. Cloud storage offers numerous advantages to users like scalability, low capital expenses, and data available from any place, anytime, regardless of location and device. However, as the users lose physical access and control over data, the storage service raises security and privacy issues, such as confidentiality, integrity, and availability of outsourced data. Data integrity is a primary concern for cloud users to confirm whether data integrity is intact or not. This paper presents a comprehensive review of cloud data auditing schemes and a comparative analysis of the desirable features. Furthermore, it provides advantages and disadvantages of the state-of-the-art techniques and a performance comparison regarding the communicational and computational costs of involved entities. It also highlights desirable features of different techniques, open issues, and future research trends of cloud data auditing protocols.

Author 1: Muhammad Farooq

Author 2: Mohd Rushdi Idrus

Author 3: Adi Affandi Ahmad

Author 4: Ahmad Hanis Mohd Shabli

Author 5: Osman Ghazali

Keywords: Cloud computing; proof of possession; data integrity auditing; proof of retrievability; public auditing; proof of ownership

Paper 44: Promises, Challenges and Opportunities of Integrating SDN and Blockchain with IoT Applications: A Survey

Abstract: Security is a major issue in the IT world, and its aim is to maintain user confidence and the coherence of the entire information system. Various international and European research projects, as well as IT manufacturers, have proposed new solutions and mechanisms to solve the problem of security in the IoT environment. Software-Defined Networking (SDN) and Blockchain are advanced technologies utilized globally for establishing secure network communication and constructing resilient network infrastructures. They serve as a robust and dependable foundation for addressing various challenges, including security, privacy, scalability, and access control. Indeed, SDN and Blockchain technologies have demonstrated their ability to efficiently manage resource utilization and facilitate secure network communication within the Internet of Things (IoT) ecosystem. Nonetheless, there exists a research gap concerning the creation of a comprehensive framework that can fulfill the unique requirements of the IoT environment. Consequently, this paper presents a recent investigation into the integration of SDN and Blockchain with IoT. The objective is to analyze their primary contributions and identify the challenges involved. Subsequently, we offer relevant recommendations to address these challenges and enhance the security and privacy of the IoT landscape.

Author 1: Loubna Elhaloui

Author 2: Mohamed Tabaa

Author 3: Sanaa Elfilali

Author 4: El habib Benlahmar

Keywords: Internet of things; SDN; blockchain

Paper 45: Application of Machine Learning in Learning Problems and Disorders: A Systematic Review

Abstract: Learning Disorders, which affect approximately 10% of the school population, represent a significant challenge in the educational field. The lack of proper diagnosis and treatment can have profound consequences, triggering psychological problems in those affected by disorders that impact reading, writing, numeracy and attention, among others. Notable among them are Attention Deficit Hyperactivity Disorder (ADHD) and dyslexia. In this context, a literature review focusing on Machine Learning applications to address these educational problems is addressed. The methodology proposed by Barbara Kitchenham guides this analysis, using the online tool Parsifal for the review, generation of search strings, formulation of research questions and management of information sources. The first findings of this research highlight a growing trend in the application of Machine Learning techniques in learning problems and disorders, especially in the last five years, as of 2019. Among the primary sources, the IEEE Digital Library emerges as a key source of information in this rapidly developing field. This innovative approach has the potential to significantly improve early detection, accurate diagnosis and implementation of personalized interventions, thus offering new perspectives in understanding and addressing the educational challenges associated with Learning Disorders.

Author 1: Mario Aquino Cruz

Author 2: Oscar Alcides Choquehuallpa Hurtado

Author 3: Esther Calatayud Madariaga

Keywords: Machine learning; learning disorder; deep learning; ADHD; dyslexia; learning impairment

Paper 46: Durian Disease Classification using Vision Transformer for Cutting-Edge Disease Control

Abstract: The durian fruit holds a prominent position as a beloved fruit not only in ASEAN countries but also in European nations. Its significant potential for contributing to economic growth in the agricultural sector is undeniable. However, the prevalence of durian leaf diseases in various ASEAN countries, including Malaysia, Indonesia, the Philippines, and Thailand, presents formidable challenges. Traditionally, the identification of these leaf diseases has relied on manual visual inspection, a laborious and time-consuming process. In response to this challenge, an innovative approach is presented for the classification and recognition of durian leaf diseases, delves into cutting-edge disease control strategies using vision transformer. The diseases include the classes of leaf spot, blight sport, algal leaf spot and healthy class. Our methodology incorporates the utilization of well-established deep learning models, specifically vision transformer model, with meticulous fine-tuning of hyperparameters such as epochs, optimizers, and maximum learning rates. Notably, our research demonstrates an outstanding achievement: vision transformer attains an impressive accuracy rate of 94.12% through the hyperparameter of the Adam optimizer with a maximum learning rate of 0.001. This work not only provides a robust solution for durian disease control but also showcases the potential of advanced deep learning techniques in agricultural practices. Our work contributes to the broader field of precision agriculture and underscores the critical role of technology in securing the future of durian farming.

Author 1: Marizuana Mat Daud

Author 2: Abdelrahman Abualqumssan

Author 3: Fadilla ‘Atyka Nor Rashid

Author 4: Mohamad Hanif Md Saad

Author 5: Wan Mimi Diyana Wan Zaki

Author 6: Nurhizam Safie Mohd Satar

Keywords: Vision transformer; durian disease; deep learning; disease control

Paper 47: The Construction of Campus Network Public Opinion Analysis Model Based on T-GAN Model

Abstract: The advancement of information technology has made the internet and social media an indispensable part of modern life, but with it comes a flood of false information and rumors. The aim of this study is to develop a technology that can automatically identify campus network public opinion information, in order to protect student groups from the intrusion of erroneous information, maintain their mental health, and promote a clear campus public opinion environment. This study used the Scrapy framework to write web scraping scripts to collect campus public opinion data, and carried out cleaning and preprocessing. Then, a transformer based generative adversarial network (T-GAN) model was designed, combined with a multi-scale convolutional neural network (MCNN) structure, for public opinion analysis on campus networks. The results show that the accuracy of the dataset processed by the T-GAN model has been improved on LGBT, KNN, SVM, and RoBERTa, proving that the campus network public opinion analysis model based on the T-GAN model helps to automatically identify campus network public opinion, protect students' physical and mental health, and promote the healthy development of the campus network environment.

Author 1: Jianan Zhang

Keywords: Public opinion analysis; T-GAN; feature extraction; multi-scale convolutional neural network; campus network

Paper 48: Method for Hyperparameter Tuning of EfficientNetV2-based Image Classification by Deliberately Modifying Optuna Tuned Result

Abstract: Method for hyperparameter tuning of EfficientNetV2-based image classification by deliberately modifying Optuna tuned result is proposed. An example of the proposed method for textile pattern quality evaluation (good or bad textile pattern fluctuation quality classification) is shown. When using the hyperparameters obtained by Optuna without changing them, the accuracy certainly improved. Furthermore, as a result of learning by changing the hyperparameter with the highest degree of importance, the accuracy changed, so it could be said that the degree of importance was certainly high. However, the accuracy also changes when learning is performed by changing the least important hyperparameter, and sometimes the accuracy is improved compared to when learning is performed using the optimal hyperparameter. From this result, it is found that the optimal hyperparameters obtained with Optuna are not necessarily optimal.

Author 1: Jin Shimazoe

Author 2: Kohei Arai

Author 3: Mari Oda

Keywords: Hyperparameter tuning; EfficientNetV2; Optuna; textile pattern; optimal hyperparameter; learning process; pattern fluctuation

Paper 49: A Novel Framework for Risk Prediction in the Health Insurance Sector using GIS and Machine Learning

Abstract: Evaluation of risk is a key component to categorize the customers of the life insurance businesses. The underwriting technique is carried out by the industries to charge the policies appropriately. Due to the availability of data hugely, the automation of underwriting process can be done using data analytics technology. Due to this, the underwriting process becomes faster and therefore quickly processes a large number of applications. This study is carried to enhance risk assessment of the applicants of life insurance industries using predictive analytics. In this research, the Geographical Information Systems (GIS) system is used to collect the data such as Air pollution, Industrial area, Covid-19 and Malaria of various geographic areas of our country, since these factors attribute to the risk of an applicant of life insurance business. Thereafter, the research is carried out using this dataset along with another dataset containing more than 50,000 entries of normal attributes of applicants of a life insurance company. Artificial Neural Network (ANN), Decision Tree (DT), and Random forest (RF) algorithms are applied on both the datasets to predict the risks of the applicants. The results showed that random forest outperformed among all the algorithms, providing the more accurate result.

Author 1: Prasanta Baruah

Author 2: Pankaj Pratap Singh

Author 3: Sanjiv kumar Ojah

Keywords: Risk prediction; data analytics; predictive analytics; underwriting; geographical information systems; random forest; artificial neural network; decision tree

Paper 50: Enhancing Underwater Object Recognition Through the Synergy of Transformer and Feature Enhancement Techniques

Abstract: Underwater object recognition presents a unique set of challenges due to the complex and dynamic characteristics of marine environments. This paper introduces a novel, multi-layered architecture that leverages the capabilities of Swin Transformer modules to process segmented image patches derived from aquatic scenes. A key component of our approach is the integration of the Feature Alignment Module (FAM), which is designed to address the complexities of underwater object recognition by enabling the model to selectively emphasize essential features. It combines multi-level features from various network stages, thereby enhancing the depth and scope of feature representation. Furthermore, this paper incorporates multiple detection heads, each embedded with the innovative ACmix module. This module offers an integrated fusion of convolution and self-attention mechanisms, refining detection precision. With the combined strengths of the Swin Transformer, FAM, and ACmix module, the proposed method achieves significant improvements in underwater object detection. To demonstrate the robustness and effectiveness of the proposed method, we conducted experiments on the UTDAC2020 dataset, highlighting its potential and contributions to the field.

Author 1: Hoanh Nguyen

Author 2: Tuan Anh Nguyen

Keywords: Underwater object recognition; swin transformer; self-attention; feature alignment

Paper 51: A Computational Prediction Model of Blood-Brain Barrier Penetration Based on Machine Learning Approaches

Abstract: Within the field of medical sciences, addressing brain illnesses such as Alzheimer's disease, Parkinson's disease, and brain tumors poses significant difficulties. Despite thorough investigation, the search for truly successful neurotherapies continues to be challenging to achieve. The blood-brain barrier (BBB), which is currently a major area of research, restricts the passage of medicinal substances into the central nervous system (CNS). It is crucial in the field of neuroscience to create drugs that can effectively cross the blood-brain barrier (BBB) and treat cognitive disorders. The objective of this study is to improve the accuracy of machine learning models in predicting BBB permeability, which is a critical factor in medication development. In recent times, a range of machine learning models such as Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Logistic Regression (LR), Artificial Neural Networks (ANN), and Random Forests (RF) have been utilized for BBB. By employing descriptors of varying dimensions (1D, 2D, or 3D), these models demonstrate the potential to make precise predictions. However, the majority of these studies are biased to the nature of datasets. To accomplish our objective, we utilized three BBB datasets for training and testing our model. The Random Forest (RF) model has shown exceptional performance when used on larger datasets and extensive feature sets. The RF model attained an overall accuracy of 90.36% with 10-fold cross-validation. Additionally, it earned an AUC of 0.96, a sensitivity of 77.73%, and a specificity of 94.74%. The assessment of an external dataset resulted in an accuracy rate of 91.89%, an AUC value of 0.94, a sensitivity rate of 91.43%, and a specificity rate of 92.31%.

Author 1: Deep Himmatbhai Ajabani

Keywords: Central Nervous System (CNS); Blood-Brain Barrier (BBB); Machine Learning (ML); Simplified Molecular Input Line Entry System (SMILES); Support Vector Machine (SVM); K-Nearest Neighbor (KNN); Logistic Regression (LR); Multi-Layer Perceptron (MLP); Light Gradient Boosting Machine (LightGBM); Random Forest (RF)

Paper 52: Enhanced Atrial Fibrillation Detection-based Wavelet Scattering Transform with Time Window Selection and Neural Network Integration

Abstract: Atrial Fibrillation (AF), a prevalent anomaly in cardiac rhythm, significantly impacts a substantial portion of the population, with projections indicating an escalation in its prevalence in the near future. This disorder manifests as irregular and accelerated heartbeats originating within the heart's upper chambers known as the atria. Neglecting to address this condition could potentially lead to serious consequences, particularly an elevated susceptibility to stroke and heart failure. This underscores the critical importance of developing an automated approach for detecting AF. In our study, an automatic approach was introduced for classifying short single-lead Electrocardiogram (ECG) recordings signals into four categories: Atrial fibrillation (AF), Normal rhythm (N), Noisy rhythm (~), or Other rhythms (O). The wavelet scattering network (WSN) is employed to extract morphological features from the ECG signals, which are then inputted into an Artificial Neural Network (ANN) with time windows selection and majority vote. The results from the testing data exhibit that our proposed model outperforms the state-of-art models, achieving a remarkable overall accuracy of 87.35% and an F1 score of 89.13%.

Author 1: Mohamed Elmehdi Ait Bourkha

Author 2: Anas Hatim

Author 3: Dounia Nasir

Author 4: Said El Beid

Keywords: Electrocardiogram (ECG); Atrial Fibrillation (AF); Wavelet Scattering Network (WSN); Artificial Neural Network (ANN)

Paper 53: Advancing Road Safety: Precision Driver Detection System with Integrated Overspeed, Alcohol Detection, and Tracking Capabilities

Abstract: In response to ongoing concerns about road accidents linked to overspeeding and drunk driving, this study introduces a groundbreaking solution: The Integrated Driver Safety system. It is a comprehensive vehicle safety system designed for real-time prevention. Crafted with cutting-edge components including ESP32, MQ3 sensor, relay, and GPS, this system operates on a dual framework. It swiftly detects instances of over speeding, triggering immediate email alerts, while concurrently inhibiting engine ignition upon detecting alcohol consumption, actively thwarting drunk driving attempts. This proactive approach not only provides real-time notifications but physically prevents intoxicated driving, drastically reducing accidents caused by these factors. With an impressive overspeed detection accuracy surpassing 95% and an efficient alcohol monitoring system, this technology cultivates responsible driving habits. Its potential widespread adoption foretells a future where road safety reaches unprecedented levels, underscoring the industry's dedication to innovation and safer driving experiences. Through this research, a compelling case emerges for the global embrace of these innovative preventive measures, illuminating a path toward significantly enhanced road safety standards.

Author 1: Jamil Abedalrahim Jamil Alsayaydeh

Author 2: Mohd Faizal bin Yusof

Author 3: Khivisha S. Mohan

Author 4: A K M Zakir Hossain

Author 5: Serhii Leoshchenko

Keywords: Integrated driver safety; overspeed detection system; alcohol monitoring technology; comprehensive vehicle security; real-time accident prevention; ESP32 GPS safety; responsible driving solutions

Paper 54: Towards a Reference Architecture for Semantic Interoperability in Multi-Cloud Platforms

Abstract: This paper focuses on semantic interoperability as one of the most significant issues in multi-cloud platforms. Organizations and individuals that adopt the multi-cloud strategy often use various cloud services and platforms. On top of that, cloud service providers may offer a range of services with unique data formats, structures, and semantics. Hence, semantic interoperability is required to enable applications and services to understand and use data consistently, regardless of the cloud service providers. The main goal of this study is to propose a reference architecture for semantic interoperability in multi-cloud platforms. Towards achieving the main goal, this paper presents two main contributions. First contribution is an extended cloud computing interoperability taxonomy, with semantic approach as one of the solutions for facilitating semantic cloud interoperability. Two fundamental semantic approaches have been identified, namely semantic technologies and frameworks which will be adopted as the main building blocks. Semantic technologies, such as ontologies, can be used to represent the semantics or meanings of data. Data may be reliably represented across multiple cloud platforms by employing a common ontology. This promotes semantic interoperability by ensuring that data is interpreted and processed uniformly within diverse cloud platforms. On the other hand, a framework offers a standardized and organized way for managing, exchanging, and representing data and services. For the second contribution of this paper, a review of recent (2018-2023) related works has been conducted by investigating the state-of-the-art of semantic interoperability in multi-cloud platforms. As a result, the proposed solution will be implemented in the context of a reference architecture. The reference architecture will act as a blueprint to systematically represent semantic interoperability in multi-cloud platforms using a hybrid approach of role-based and layer-based. Additionally, a semantic layer will be extended to the reference architecture to facilitate semantic interoperability.

Author 1: Norazian M Hamdan

Author 2: Novia Admodisastro

Keywords: Cloud computing; multi-cloud; reference architecture; semantic interoperability; semantic technologies

Paper 55: Pancreatic Cancer Detection Through Hyperparameter Tuning and Ensemble Methods

Abstract: Computing techniques have brought about a significant transformation in the field of medical research. Machine learning techniques have facilitated the analysis of vast amounts of data, modeling of complex scenarios, and the ability to make well-informed decisions. This presents an opportunity to develop reliable and effective medical system implementations, which may include the automatic recognition of uncertain issues related to health. Currently, significant research efforts to be directed towards the prediction of cancer, particularly focusing on addressing the various health complications caused by this disease, which can adversely impact multiple organs within the body. Pancreatic Cancer (PC) stands out as a highly lethal form of tumor, with a rather discouraging global five-year survival rate of approximately 5%. The truth behind the early detection increases the survival rate and it also helps the radiologists to give better treatment to those who are affected at early stages. Creatinine, LYVE1, REG1B, and TFF1 are urine proteomic biomarkers that offer a promising non-invasive and affordable diagnostic technique for detecting pancreatic cancer. In this study, a novel model that combines gridsearchCV technique to search and find the optimal combination of hyperparameters for a random forest classifier. In this research a new ensemble method to enhance the performance for classification of pancreatic cancer and non-cancer by using urinary biomarkers which is collected from Kaggle. The implemented model achieved better results of Accuracy 99.98%, F-1 score 99.98, Precision 99.98, and Recall 99.98.

Author 1: Koteswaramma Dodda

Author 2: G. Muneeswari

Keywords: Pancreatic cancer; Machine learning (ML); urinary biomarkers; grid search hyper parameter tuning; Random Forest (RF)

Paper 56: An Efficient Honeycomb Lung Segmentation Network Combining Multi-Paradigms Representation and Cascade Attention

Abstract: Honeycomb lung is a pulmonary manifestation that occurs in the terminal stage of various lung diseases, which greatly threatens patients. Due to the different locations and irregular shapes of lesions, the accurate segmentation of the honeycomb region is an essential and challenging problem. However, most deep learning methods struggle to effectively utilize both global and local information from lesion images, resulting in cannot to accurately segment the lesion. In addition, these methods often ignore some semantic information that is necessary for the segmentation of lesion location and shape in the decoding stage. To alleviate these challenges, in this paper, we propose a dual-branch encoder and cascaded decoder network (DECDNet) for segmenting honeycombs lesions. First, we design a dual-branch encoder consisting of ResNet34 and Swin-Transformer with different paradigm representations to extract local features and long-range dependencies respectively. Next, to further combine the different paradigm features, we develop the feature fusion module to obtain richer representation information. Finally, considering the problem of information loss during the decoder, a cascaded attention decoder is constructed to aggregate the multi-stage encoder information to get the final segmentation result. Experimental results demonstrate that our method outperforms other methods on the in-house honeycomb lung dataset. Notably, compared with the other nine universal methods, the proposed DECDNet obtains the highest IoU (86.34%), Dice (92.66%), Precision (93.21%), Recall (92.13%), F1-Score (92.66%), and achieves the lowest HD95 (7.33) and ASD (2.30). In particular, our method enables precisely segmenting lesions under different clinical scenarios as well. Our code and dataset are available at https://github.com/ybq17/DECDNet.

Author 1: Bingqian Yang

Author 2: Xiufang Feng

Author 3: Yunyun Dong

Keywords: Honeycomb lung; attention; convolutional neural network; transformer; image segmentation

Paper 57: Efficient Deep Reinforcement Learning for Smart Buildings: Integrating Energy Storage Systems Through Advanced Energy Management Strategies

Abstract: This study presents a novel and workable approach to solving the critical issue of improving energy management in smart buildings. Using a large dataset from a seven-story office building in Bangkok, Thailand, our work introduces a novel approach that combines Deep Q-network (DQN) algorithms with energy storage models and cost optimization strategies. The suggested approach is intended to reduce operational expenses, improve the energy economic performance, and efficiently control peak demand. The energy storage model used in this research incorporates the use of the capabilities of advanced storage models in smart buildings, particularly lithium-ion batteries and supercapacitors. When the cost optimization approach is applied using linear programming, energy consumption costs are significantly reduced. Notably, our method outperforms current algorithms, specifically outperforming them, to show its effectiveness in smart building energy management by outperforming current algorithms, especially Genetic and Fuzzy Algorithms. In comparison to traditional methods, the DQN algorithm exhibits an impressive 8.6% reduction in Mean Square Error (MSE) and a 6.4% drop in Mean Absolute Error (MAE), making it a standout performer in the research through Python software. The results highlight the significance of optimizing DQN algorithm parameters for best outcomes, with a focus on adaptability to various properties of smart buildings. This investigation is novel because it integrates cost optimization, reinforcement learning, and energy storage. This results in a flexible and all-inclusive framework that can be used for effective and sustainable energy management in smart buildings.

Author 1: Artika Farhana

Author 2: Nimmati Satheesh

Author 3: Ramya M

Author 4: Janjhyam Venkata Naga Ramesh

Author 5: Yousef A. Baker El-Ebiary

Keywords: Deep q-network; cost optimization; smart building; energy management; peak demand

Paper 58: Use of ANN, LSTM and CNN Classifiers for the New MSCC and BSCC Methods in the Detection of Parkinson's Disease by Voice Analysis

Abstract: Parkinson's disease (PD) is a neurodegenerative condition that impacts a significant global population. The timely and precise identification of PD plays a pivotal role in facilitating early intervention and the efficient management of the condition. Recently, speech analysis has emerged as a promising non-invasive technique for the detection of PD due to its accessibility and ability to reveal subtle vocal biomarkers associated with the disease. This research introduces an innovative approach utilizing Short-Time Fourier Transform (STFT) to generate spectrograms, specifically Bark Spectrogram Cepstral Coefficients (BSCC) and Mel Spectrogram Cepstral Coefficients (MSCC). These coefficients are compared with traditional and well-known coefficients, namely Mel-Frequency Cepstral Coefficients (MFCC) and Bark Frequency Cepstral Coefficients (BFCC). To extract the most effective coefficients for Parkinson's disease detection, three robust classification techniques—Long Short-Term Memory neural networks (LSTM), Convolutional Neural Networks (CNN), and Artificial Neural Networks (ANN)—are employed. As a result, the BSCC and MSCC algorithms achieve a maximum accuracy rate of 90%, surpassing the accuracy of the traditional MFCC and BFCC coefficients. Therefore, these newly proposed coefficients prove to be more precise in diagnosing Parkinson's disease compared to the conventional MFCC and BFCC coefficients.

Author 1: Miyara Mounia

Author 2: Boualoulou Nouhaila

Author 3: Nsiri Benayad

Author 4: Belhoussine Drissi Taoufiq

Keywords: Parkinson’s Disease (PD); Bark Spectrogram Cepstral Coefficients (BSCC); Mel Spectrogram Cepstral Coefficients (MSCC); Long-Term Memory Neural Networks (LSTM); Convolutional Neural Networks (CNN); Artificial Neural Networks (ANN)

Paper 59: Artificial Intelligence for Confidential Information Sharing Based on Knowledge-Based System

Abstract: Ensuring the security of sensitive data and protecting user privacy remains one of the most significant challenges in our contemporary landscape. Organizations Companies cannot adopt a new technology without reassurance regarding data confidentiality. To address these challenges, we present an innovative system that draws upon extensive knowledge and expertise in the field of cryptography, especially in encryption methods. This system tailors its strategies to align with specific scenarios, prioritizing data confidentiality. Our solution is based on one of the Artificial Intelligence techniques, which is Knowledge-Based Systems (KBS) and extends the intelligent encryption methods from our previous research. However, this new system has taken a novel approach by reconfiguring this within KBS architecture. We have introduced additional technical components, including knowledge bases, an inference engine, and the Nearest Neighbor (NN) search algorithm. As a result, this revised architecture not only enhances security and system performance but also showcases improved maintainability and scalability.

Author 1: Bouchra Boulahiat

Author 2: Salima Trichni

Author 3: Mohammed Bougrine

Author 4: Fouzia Omary

Keywords: IT security; cryptography; confidentiality; Knowledge-Based system; artificial intelligence

Paper 60: Enhancing Software User Interface Testing Through Few Shot Deep Learning: A Novel Approach for Automated Accuracy and Usability Evaluation

Abstract: Traditional user interface (UI) testing methods in software development are time-consuming and prone to human error, requiring more efficient and accurate approaches. Moreover, deep learning requires extensive data training to develop accurate automated UI software testing. This paper proposes an efficient and accurate method for automating UI software testing using Deep learning with training data limitations. We propose a novel deep learning-based framework suitable for UI element analysis in data-scarce situations, focusing on Few-shot learning. Our framework initiates with several robust feature extraction modules that employ and compare sophisticated encoder models to be adept at capturing complex patterns from a sparse dataset. The methodology employs the Enrico and UI screen mistake datasets, overcoming training data limitations. Utilizing encoder models, including CNN, VGG-16, ResNet-50, MobileNet-V3, and EfficientNet-B1, the EfficientNet-B1 model excelled in the setting of Few-Shot learning with five-shot with an average accuracy of 76.05%. Our proposed model's accuracy was improved and compared to the state-of-the-art method. Our findings demonstrate the effectiveness of few-shot learning in UI screen classification, setting new benchmarks in software testing and usability evaluation, particularly in limited data scenarios.

Author 1: Aris Puji Widodo

Author 2: Adi Wibowo

Author 3: Kabul Kurniawan

Keywords: Deep learning; efficientnet; few-shot; software testing; UI screen classification

Paper 61: Optimizing Crop Yield Prediction in Precision Agriculture with Hyperspectral Imaging-Unmixing and Deep Learning

Abstract: The optimization of crop yield projections has arisen as a major problem in modern agriculture, due to the increasing demand for food supply and the necessity for effective resource management. Precision and scalability are hampered by the limits associated with conventional agricultural production prediction techniques, which mostly rely on observations and simple data sources. While methods like random forest (RF) and K-nearest neighbors (KNN) are widely used, their reliance on personal assessments and insufficient knowledge of crop attributes typically results in less accurate forecasts and makes them unsuitable for agricultural precision. The suggested method combines deep learning, spectral unmixing, and hyperspectral imaging methods to overcome these obstacles. With the use of hyperspectral imaging, which records a vast array of data that is not visible to the human eye, crop attributes may be thoroughly examined and can identify the unique spectral fingerprints of different agricultural constituents by using spectral unmixing approaches, which makes it easier to evaluate the health and growth phases of the crop. Then, using this augmented spectral data, deep learning algorithms create a solid, data-driven basis for precise crop production prediction. MATLAB has been used in the suggested workflow. The combination of deep learning, spectrum unmixing, and hyperspectral imaging provides a comprehensive, cutting-edge approach that goes beyond the constraints of conventional techniques were implemented in python. Some of the algorithms that were examined, this one with integration has the lowest Root Mean Square Error (RMSE) of 0.15 and Mean Absolute Error (MAE) of 0.14, demonstrating higher prediction accuracy above other current models. This novel method represents a substantial breakthrough in precision agriculture while also improving crop production prediction.

Author 1: Deeba K

Author 2: O. Rama Devi

Author 3: Mohammed Saleh Al Ansari

Author 4: Bhargavi Peddi Reddy

Author 5: Manohara H T

Author 6: Yousef A. Baker El-Ebiary

Author 7: Manikandan Rengarajan

Keywords: Crop yield prediction; hyper spectral image; spectral unmixing; resource management; precision agriculture

Paper 62: Optimizing Network Security and Performance Through the Integration of Hybrid GAN-RNN Models in SDN-based Access Control and Traffic Engineering

Abstract: By offering flexible and adaptable infrastructures Software-Defined Networking (SDN) has emerged as a disruptive technology that has completely changed network provisioning and administration. By seamlessly integrating Hybrid Generative Adversarial Network-Recurrent Neural Network (GAN-RNN) modeling into the foundation of SDN-based traffic engineering and accessibility control methods, this work presents a novel and comprehensive method to improve network efficiency and security. The proposed Hybrid GAN-RNN models address two important aspects of network management: traffic optimization and access control. They combine the benefits of Generative Adversarial Networks (GANs) and Recurrent Neural Networks (RNNs). Traditional traffic engineering techniques frequently find it difficult to quickly adjust to situations that are changing quickly within today's dynamic networking environments. The models' capacity to generate synthetic traffic patterns that nearly perfectly replicate the complexity of real network traffic demonstrates the power of GANs. Network administrators can now allocate resources and routing methods more dynamically, as well as in responding to real-time network inconsistencies, due to this state-of-the-art technology. The technique known as Hybrid GAN-RNN addresses the enduring problem of network security. With their reputation for continuous learning and by utilizing Python software, recurrent neural networks (RNNs) are at the forefront of developing flexible management of access rules. With an incredible 99.4% accuracy rate, the "Proposed GAN-RNN" approach outperforms the other approaches. A comprehensive evaluation of network traffic and new safety risks allow for the immediate modification of these policies. This work is interesting because it combines hybrid GAN-RNN algorithms to strengthen security protocols with adaptive access control while also optimizing network efficiency through realistic traffic modeling.

Author 1: Ganesh Khekare

Author 2: K. Pavan Kumar

Author 3: Kundeti Naga Prasanthi

Author 4: Sanjiv Rao Godla

Author 5: Venubabu Rachapudi

Author 6: Mohammed Saleh Al Ansari

Author 7: Yousef A. Baker El-Ebiary

Keywords: Software-defined networking; generative adversarial networks; recurrent neural networks; traffic engineering

Paper 63: The Contribution of Health Management Information Systems to Enhancing Healthcare Operations

Abstract: Various strategies for enhancing quality have been implemented by developed and developing countries in light of the worldwide emphasis on bolstering healthcare systems. Many nations are currently directing their attention towards bolstering their existing information systems or establishing new ones, recognizing the critical role of information in the functioning of healthcare systems. The study aimed to assess the impact of leadership style, organizational factors, technology, and healthcare provider behavior on the implementation of health management information systems in healthcare organizations. While the study was informed by the performance framework of routine information systems, it was primarily based on system theory. After conducting the analysis in Python and SPSS, the data was presented using descriptive statistics, such as means and standard deviations, and inferential statistics including regression analysis. The study observed that information timelines significantly moderated the connection between the technical factor and the integration of health management information systems.

Author 1: Majzoob K. Omer

Keywords: Health management information system; cronbach alpha; moderator; information timelines

Paper 64: A Sophisticated Deep Learning Framework of Advanced Techniques to Detect Malicious Users in Online Social Networks

Abstract: Malicious user detection is a cybersecurity exploration domain because of the emergent jeopardies of data breaches and cyberattacks. Malicious users have the potential to detriment the system by engaging in unauthorized actions or thieving sensitive data. This paper proposes the dual-powered CLM technique (Convolution neural networks and LSTM) and optimization technique, a sophisticated methodology for distinguishing malicious user behavior that assimilates LSTM and CNN, and finally optimization technique to enhance the results. A genetic algorithm is used to augment the model's capability to perceive altering and nuanced malicious performance by fine-tuning its parameters. Due to the rising vulnerabilities of data breaches and cyber-attacks, malicious user identification in OSN (Online Social Networks) is a significant topic of research in cybersecurity. The proposed technique pursues to ascertain anomalous user behavior patterns by assessing vast quantities of data generated by digital systems with CLM and optimizing detection accuracy with genetic algorithms. On a public dataset of social media bot dataset, a twibot-20 dataset comprehending user activity data, was explored to measure the performance of the suggested methodology. The outcomes demonstrated that, in comparison to conventional machine learning algorithms like SVM and RF, which respectively obtained 92.3% and 88.9% accuracy, our technique, had a better accuracy of 98.7%. Moreover, the other metrics measures were assessed, and the proposed technique outperformed traditional machine learning algorithms in each situation.

Author 1: Sailaja Terumalasetti

Author 2: Reeja S R

Keywords: Online social networks; malicious user behavior; convolution neural networks; long short-term memory; genetic algorithm

Paper 65: Exploring a Novel Machine Learning Approach for Evaluating Parkinson's Disease, Duration, and Vitamin D Level

Abstract: Parkinson's disease is an increasingly prevalent, degenerative neurological condition predominantly afflicting individuals aged 50 and older. As global life expectancy continues to rise, the imperative for a deeper comprehension of factors influencing the course and intensity of PD becomes more pronounced. This investigation delves into these facets, scrutinizing various parameters including patient medical history, dietary practices, and vitamin D levels. A dataset comprising 50 PD patients and 50 healthy controls, sourced from Dhaka Medical Institute, serves as the foundation for this study. Machine learning techniques, notably the Modified Random Forest Classifier (MRFC), are harnessed to prognosticate both PD severity and duration. Strikingly, the MRFC-based prediction model for PD severity attains an impressive accuracy of 97.14%, while the predictive model for PD duration demonstrates an accuracy of 95.16%. Noteworthy is the observation that vitamin D levels are notably higher in the healthy cohort compared to PD-afflicted individuals, exerting a substantial positive influence on both the severity and duration predictions, surpassing the influence of other measured parameters. This inquiry underscores the practicality of machine learning in forecasting PD progression and duration and underscores the pivotal role of vitamin D levels as a predictive factor. These discoveries provide invaluable insights into advancing our comprehension and management of PD in an aging population.

Author 1: Md. Asraf Ali

Author 2: Md. Kishor Morol

Author 3: Muhammad F Mridha

Author 4: Nafiz Fahad

Author 5: Md Sadi Al Huda

Author 6: Nasim Ahmed

Keywords: Parkinson's disease; machine learning; vitamin D; severity; disease duration

Paper 66: Applying Big Data Analysis and Machine Learning Approaches for Optimal Production Management

Abstract: In this research paper, we delve into the transformative potential of integrating Big Data analytics with machine learning (ML) techniques, orchestrating a paradigm shift in production management methodologies. Traditional production systems, often marred by inefficiencies stemming from data opacity, have encountered bottlenecks that throttle scalability and adaptability, particularly in complex, fluctuating markets. By harnessing the voluminous streams of data—both structured and unstructured—generated in contemporary production environments, and subjecting these data lakes to advanced ML algorithms, we unveil profound insights and predictive patterns that remain elusive under conventional analytical methods. Our discourse juxtaposes the multidimensionality of Big Data—emphasizing velocity, variety, veracity, and volume—with the finesse of ML models, such as neural networks and reinforcement learning, which adapt iteratively to the dynamism inherent in production landscapes. This symbiosis underpins a more holistic, anticipatory decision-making process, empowering stakeholders to pinpoint and mitigate operational hiccups, optimize supply chain vectors, and streamline quality assurance protocols, thereby catalyzing a more resilient, responsive, and cost-effective production framework. Furthermore, we explore the ethical contours of data stewardship in this context, advocating for a judicious balance between technological ascendancy and responsible data governance. The culmination of this exploration is the conceptualization of a predictive, self-regulating production ecosystem that thrives on continuous learning and improvement, dynamically calibrating itself in response to an ever-evolving market tableau and thereby heralding a new era of optimal, sustainable, and intelligent production management.

Author 1: Sarsenkul Tileubay

Author 2: Bayanali Doszhanov

Author 3: Bulgyn Mailykhanova

Author 4: Nurlan Kulmurzayev

Author 5: Aisanim Sarsenbayeva

Author 6: Zhadyra Akanova

Author 7: Sveta Toxanova

Keywords: Optimal production; smart manufacturing; machine learning; big data; management

Paper 67: Development of an Intelligent Service Delivery System to Increase Efficiency of Software Defined Networks

Abstract: The burgeoning complexity in network management has garnered considerable attention, specifically focusing on Software-Defined Networking (SDN), a transformative technology that addresses limitations inherent in traditional network infrastructures. Despite its advantages, SDN is often susceptible to bottlenecks and excessive load issues, underscoring the necessity for more robust load balancing solutions. Previous research in this realm has predominantly concentrated on employing static or dynamic methodologies, encapsulating only a handful of parameters for traffic management, thereby limiting their effectiveness. This study introduces an innovative, intelligence-led approach to service delivery systems in SDN, specifically by orchestrating packet forwarding—encompassing both TCP and UDP traffic—through a multi-faceted analysis utilizing twelve distinct parameters elaborated in subsequent sections. This research leverages advanced machine learning algorithms, notably K-Means and DBSCAN clustering, to discern patterns and optimize traffic distribution, ensuring a more nuanced, responsive load balancing mechanism. A salient feature of this methodology involves determining the ideal number of operational clusters to enhance efficiency systematically. The proposed system underwent rigorous testing with an escalating scale of network packets, encompassing counts of 5,000 to an extensive 10,000,000, to validate performance under varying load conditions. Comparative analysis between K-Means and DBSCAN's results reveals critical insights into their operational efficacy, corroborated by juxtaposition with extant scholarly perspectives. This investigation's findings significantly contribute to the discourse on adaptive network solutions, demonstrating that an intelligent, parameter-rich approach can substantively mitigate load-related challenges, thereby revolutionizing service delivery paradigms within Software-Defined Networks.

Author 1: Serik Joldasbayev

Author 2: Saya Sapakova

Author 3: Almash Zhaksylyk

Author 4: Bakhytzhan Kulambayev

Author 5: Reanta Armankyzy

Author 6: Aruzhan Bolysbek

Keywords: Load balancing; machine learning; server; classification; software

Paper 68: A Comprehensive Review of Healthcare Prediction using Data Science with Deep Learning

Abstract: Data science in healthcare prediction technology can identify diseases and spot even the smallest changes in the patient's health factors and prevent the diseases. Several factors make data science crucial to healthcare today the most important among them is the competitive demand for valuable information in the healthcare systems. The data science technology along with Deep Learning (DL) techniques creates medical records, disease diagnosis, and especially, real-time monitoring of patients. Each DL algorithm performs differently using different datasets. The impacts on different predictive results may be affects overall results. The variability of prognostic results is large in the clinical decision-making process. Consequently, it is necessary to understand the several DL algorithms required for handling big amount of data in healthcare sector. Therefore, this review paper highlights the basic DL algorithms used for prediction, classification and explains how they are used in the healthcare sector. The goal of this review is to provide a clear overview of data science technologies in healthcare solutions. The analysis determines that each DL algorithm have several negativities. The optimal method is necessary for critical healthcare prediction data. This review also offers several examples of data science and DL to diagnose upcoming trends on the healthcare system.

Author 1: Asha Latha Thandu

Author 2: Pradeepini Gera

Keywords: Data science; deep belief network; healthcare; sparse auto encoder; deep learning

Paper 69: Enhancing Autism Severity Classification: Integrating LSTM into CNNs for Multisite Meltdown Grading

Abstract: Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by deficits in social interaction, verbal and non-verbal communication, and is often associated with cognitive and neurobehavioral challenges. Timely screening and diagnosis of ASD are crucial for early educational planning, treatment, family support, and timely medical intervention. Manual diagnostic methods are time-consuming and labor-intensive, underscoring the need for automated approaches to assist caretakers and parents. While various researchers have employed machine learning and deep learning techniques for ASD diagnosis, existing models often fall short in capturing the complexity of multisite meltdowns and fully leveraging the interdependence among these meltdowns for severity assessment in acquired facial images of children, hindering the development of a comprehensive grading system. This paper introduces a novel approach using a Long Short Term Memory (LSTM) integrated Convolution Neural Network (CNN) designed to identify multisite meltdowns and exploit their interdependence for severity assessment in ASD. The process begins with image pre-processing, involving discrete convolution filters for noise removal and contrast enhancement to improve image quality. The enhanced image then undergoes instance segmentation using the Segment Anything model to identify significant regions in the child's image. The segmented region is subjected to principal component analysis for feature extraction, and these features are utilized by the LSTM-integrated CNN for meltdown detection and severity classification. The model is trained using children's images extracted from videos, and testing is performed on videos captured during children's observations. Performance analysis reveals superior results, with a training accuracy of 88% and validation accuracy of 84%, outperforming conventional methods. This innovative approach not only enhances the efficiency of ASD diagnosis but also provides a more nuanced understanding of multisite meltdowns and their impact on severity, contributing to the development of a robust grading system.

Author 1: Sumbul Alam

Author 2: S Pravinth Raja

Author 3: Yonis Gulzar

Author 4: Mohammad Shuaib Mir

Keywords: Autism spectrum disorder; mutilating meltdown; convolution neural network; long short term memory; multisite meltdown; video classification; image classification

Paper 70: Analysis of Synthetic Data Utilization with Generative Adversarial Network in Flood Classification using K-Nearest Neighbor Algorithm

Abstract: Indonesia is a country with a tropical climate that has high rainfall rates and is supported by the uncertainty of weather and climate conditions. With the uncertainty of weather and climate as well as flood events, minimal predictive information on flooding, and the lack of availability of data on the causes of flooding, a comparison of synthetic data generation from the minimal data available from BMKG with synthetic data generation from Kaggle online platform data in the form of temperature and humidity data, rainfall, and wind speed from BMKG and annual rain data from Kaggle was analyzed. This research aims to obtain the results of data comparison analysis of synthetic data generation from different datasets with the benchmark of classification system results using K-Nearest Neighbor (KNN) and accuracy evaluation with Confusion Matrix. The research process uses climate data from the BMKG DI Yogyakarta Climatology Station within 20 months, the Geophysical Station within 12 months, and Kerala data with a range of 1901–2018. Synthetic data generation is done using the Conditional Tabular Generative Adversarial Network (CTGAN) model. CTGAN produces quite good data in terms of distribution and data differences if the original data is large and the synthetic data generated is small. The KNN classification system on the BMKG data experienced overfitting, as indicated by the accuracy value in the evaluation increasing in the range of 85–94% and the validation decreasing in the range of 89%–65%. This is because there is no uniqueness in the data and too little original data made into synthetics, which affects the difficulty of the classification system in identifying data that is quite different in distance and data values generated by CTGAN. In Kerala, the accuracy value on evaluation is in the range of 92–95%, and validation is in the range of 0.7–0.83%, with Classifier k1 being the most optimal system.

Author 1: Wahyu Afriza

Author 2: Mardhani Riasetiawan

Author 3: Dyah Aruming Tyas

Keywords: Classification; rainfall; synthetic data; KNN; GAN

Paper 71: A Piano Single Tone Recognition and Classification Method Based on CNN Model

Abstract: In order to improve the recognition and classification effect of piano single tone, this paper combines the CNN (Convolutional Neural Networks) model to construct the piano single tone recognition and classification model, and equalizes the uniformly irradiated parabolic tone transmission hardware. In this paper, the analytic method is used to calculate the direction diagram of the tone transmission hardware, and the analytical expression for calculating the gain of the tone transmission hardware is obtained. Moreover, this paper gives the calculation and analytical expression of the hardware gain of the tone transmission in the main lobe, and obtains the calculation method of the relative position of the two tone transmission hardware by using the conversion relationship between the global coordinate and the local coordinate. Finally, the variation law of the received power with the azimuth/elevation angle of the receiving tone transmission hardware and the incident high-power microwave frequency is given. The experimental study shows that the piano single tone recognition and classification method based on CNN model proposed in this paper can play an important role in piano single tone recognition. This article improves the note recognition algorithm for piano music by combining note features with frequency spectrum to obtain note spectrum, which improves the accuracy of audio classification recognition.

Author 1: Miaoping Geng

Author 2: Ruidi He

Author 3: Ziyihe Zhou

Keywords: CNN model; piano; single tone recognition; classification

Paper 72: Intelligent Evaluation and Optimization of Postgraduate Education Comprehensive Ability Training under the Mode of “One Case, Three Systems”

Abstract: This study aims to explore the intelligent evaluation and optimization methods for the comprehensive ability training of graduate students under the mode of "one case, three systems" to improve the quality and effect of graduate training. Firstly, a weighted clustering algorithm for mixed attributes is designed. Secondly, an evaluation model of postgraduate training quality based on sampling method and ensemble learning is established. Finally, the algorithm's and the model's performance are compared and tested. The test results show that with the increase in the number of experiments, the accuracy of the proposed weighted clustering algorithm can reach more than 90%, which is improved by 10%. The average number of iterations is 276, and the accuracy and F1 value can achieve the highest level with fewer iterations and stable algorithm performance. Compared with the R1 model, F1 and the accuracy of the model proposed in this study are enhanced by 3.29% and 6.75%, respectively. The feature-weighted clustering algorithm and the training quality evaluation model designed here complement each other and jointly construct a more elaborate and comprehensive training system. The feature-weighted clustering algorithm oriented to mixed attributes for the first time combines sampling methods and ensemble learning in the education ability training. Moreover, a multi-dimensional and intelligent postgraduate training evaluation framework is constructed, which provides a new idea for improving the quality of postgraduate training.

Author 1: Yong Xiang

Author 2: Zeyou Chen

Author 3: Liyu Lu

Author 4: Yao Wei

Keywords: Comprehensive ability training of graduate students; one case; three systems; feature weighted clustering algorithm; sampling method; ensemble learning

Paper 73: A Novel Fusion Deep Learning Approach for Retinal Disease Diagnosis Enhanced by Web Application Predictive Tool

Abstract: Retinal disorders such as age-related macular degeneration and diabetic macular edema can lead to permanent blindness. Optical coherence tomography (OCT) enables professionals to observe cross-sections of the retina, which aids in diagnosis. Manually analyzing images is time-consuming, difficult, and prone to mistakes. In the dynamic and constantly evolving domain of artificial intelligence (AI) and medical imaging, our research represents a significant development in the field of retinal diagnostics. In this study, we introduced “RetiNet”, an advanced hybrid model that is derived from the best features of ResNet50 and DenseNet121. To the model, we utilized an open-source retinal dataset that underwent a meticulous refinement process using a series of preprocessing techniques. The techniques involved Histogram Equalization for the purpose of achieving optimal contrast, Gaussian blur to mitigate noise, morphological operations to facilitate precise feature extraction, and Data Balancing to ensure impartial model training. These operations led to the attainment of a test accuracy of 98.50% by RetiNet, surpassing the performance standard set by existing models. A web application has been developed with the purpose of disease prediction, providing doctors with assistance in their diagnostic procedures. Through the development of RetiNet, our research not only transforms the accuracy of retinal diagnostics but also introduces an innovative combination of deep learning and application-oriented solutions. This innovation brings in a novel era characterized by improving reliability and efficiency in the field of medical diagnostics.

Author 1: Nani Gopal Barai

Author 2: Subrata Banik

Author 3: F M Javed Mehedi Shamrat

Keywords: Retinal disease; RetiNet; hybrid model; learning; Web application; gaussian blur; histogram equalization

Paper 74: Enhancing Production System Performance: Failure Detection and Availability Improvement with Deep Learning and Genetic Algorithm

Abstract: A crucial component of industrial operations is the detection of production system failures, which aims to spot any problems before they get worse. By applying cutting-edge methods like deep learning and genetic algorithms, failure detection accuracy may be improved, allowing for preemptive actions to reduce downtime and maximize system availability. These methods improve reactivity to possible errors and solve dynamic issues, which enhances the overall efficiency and reliability of production systems. This study offers a novel method for improving the availability and failure detection of production systems using deep learning techniques and genetic algorithms in a data-driven strategy. The goal of the project is to provide a complete framework for efficient failure detection that incorporates deep learning models, particularly Convolutional Neural Network (CNN) Autoencoder. Furthermore, system configurations are optimized through the use of genetic algorithms, improving overall availability. The suggested model is able to identify complex patterns and connections in the data by being trained on a variety of datasets that contain information about equipment failure. The incorporation of genetic algorithm guarantees flexibility and resilience in system setups, hence augmenting total availability. The study presents a proactive and flexible approach to the dynamic issues encountered in industrial environments, providing a notable breakthrough in failure detection and availability improvement. The proposed model is implemented in Python software. It achieves an astounding 99.32% accuracy rate, which is 3.58% higher than that of current techniques like CNN-LSTM (Long Short-Term Memory), Bi-LSTM (Bi-directional Long Short-Term Memory), and CNN-RNN (Recurrent Neural Network). The data-driven approach's high accuracy highlights its efficacy in forecasting and avoiding problems, which minimizes downtime and maximizes production efficiency.

Author 1: Artika Farhana

Author 2: Shaista Sabeer

Author 3: Ayasha Siddiqua

Author 4: Afsana Anjum

Keywords: Autoencoder; availability enhancement; convolutional neural network; failure detection; genetic algorithm

Paper 75: Enhanced Multi-Object Detection via the Integration of PSO, Kalman Filtering, and CNN Compressive Sensing

Abstract: Many inventive techniques have been created in the field of machine vision to solve the challenging challenge of detecting and tracking one or more objects in the face of challenging conditions, such as obstacles, object motion, changes in light, shaking, and rotations. This research article provides a novel method that combines Convolutional Neural Networks (CNNs), Compressive Sensing, Kalman Filtering, and Particle Swarm Optimization (PSO) to address the challenges of multi-object tracking under dynamic conditions. Initially, a CNN-based object classification and identification system is demonstrated, which efficiently locates objects in video frames. Subsequently, in order to produce precise representations of object appearances, utilizing compressive sensing techniques. The Kalman Filter ensures adaptability to irregular observations, eliminates erroneous data, and reduces uncertainty. PSO enhances tracking efficiency by optimizing forecast precision. When combined, these techniques provide robust tracking even in the presence of complex movement patterns, occlusions, and visual disparities. The efficiency of this strategy is demonstrated by an empirical investigation that produces a remarkable tracking accuracy of 98%, which is 3.15% greater than other methods across a range of challenging settings. This technique has been compared to various existing approaches, including the Clustering Method, YOLOV4 DNN Model, and YOLOV3 Model, and its deployment is made easier with Python software. This hybrid technique, which addresses the limitations of separate approaches and offers a holistic approach to multi-object monitoring, has potential applications in surveillance, robotics, and autonomous systems.

Author 1: S. V. Suresh Babu Matla

Author 2: S. Ravi

Author 3: Muralikrishna Puttagunta

Keywords: Multi-Object tracking; object detection; convolutional neural networks; kalman filtering; particle swarm optimization

Paper 76: Comparative Analysis of Weighted Ensemble and Majority Voting Algorithms for Intrusion Detection in OpenStack Cloud Environments

Abstract: In the ever-evolving landscape of cybersecurity, the detection of malicious activities within cloud environments remains a critical challenge. This research aims to compare the effectiveness of two ensemble algorithms, the weighted ensemble algorithm and the majority voting algorithm, in the context of intrusion detection within an OpenStack cloud environment. To conduct this study, a dataset was generated using a network of 10 virtual machines, simulating the complex dynamics of a real cloud infrastructure. Various attack scenarios were simulated, and system metrics including CPU usage, memory utilization, and network traffic were monitored and logged. The weighted ensemble algorithm combines the predictions of multiple individual models with varying weights, while the majority voting algorithm aggregates predictions from multiple models. Through a rigorous experimental setup, these algorithms were applied to the generated dataset, and their performance was evaluated using standard metrics such as accuracy, precision, recall, and F1-score. These findings provide valuable insights into the strengths and weaknesses of ensemble algorithms for intrusion detection in cloud environments. It highlights the importance of selecting appropriate algorithms based on specific security requirements and threat profiles. Different attack scenarios may require different algorithmic approaches to achieve optimal results. Overall, this study contributes to the understanding of ensemble techniques in cloud security and offers a foundation for further research in optimizing intrusion detection strategies within dynamic and complex cloud environments. By identifying the strengths and weaknesses of different ensemble algorithms, cybersecurity professionals can make informed decisions in selecting the most suitable approach to enhance the security of cloud environments.

Author 1: Pravin Patil

Author 2: Geetanjali Kale

Author 3: Nidhi Bivalkar

Author 4: Agneya Kolhatkar

Keywords: Intrusion detection; ensemble algorithms; cloud security; openstack; weighted ensemble; majority voting

Paper 77: Rural Homestay Spatial Planning and Design Based on Bert BiLSTM EIC Algorithm in the Background of Digital Ecology

Abstract: There is a promising development prospect in the digital ecosystem. In this context, the spatial planning of rural homestays has also received widespread attention. The research aims to better explore the advantages and determine the development direction of rural homestays, while providing two-way demand support for consumers and managers. Therefore, this study combines bidirectional long-term and short-term memory networks, pre-trained models, and emotional information attention mechanisms in deep learning. A new emotional analysis model is proposed. Then it is applied to the spatial planning of homestays near Chengdu Normal University and Chengdu Neusoft University. The experimental results show that the accuracy, recall, and F1 values of the new emotion analysis model proposed in this research reach 94%, 93%, and 94%, respectively. In terms of consumer satisfaction with the spatial location of homestays before and after the renovation, the average score of homestays near Chengdu Normal University increases by 21% compared to before the renovation. The average score of homestays near Chengdu Neusoft University increases by 40% compared to before the renovation. In summary, the new emotional analysis model proposed in this research has certain feasibility and effectiveness in the planning of rural homestay spatial location, providing new ideas for homestay spatial location planning.

Author 1: Zhibin Qiu

Author 2: Junghoon Mok

Keywords: Rural homestay; spatial planning; deep learning; emotional analysis; bidirectional long short term memory network

Paper 78: Robot Human-Machine Interaction Method Based on Natural Language Processing and Speech Recognition

Abstract: With the rapid development of artificial intelligence technology, robots have gradually entered people's lives and work. The robot human-machine interaction system for image recognition has been widely used. However, there are still many problems with robot human-machine interaction methods that utilize natural language processing and speech recognition. Therefore, this study proposes a new robot human-machine interaction method that combines structured perceptron lexical analysis model and transfer dependency syntactic analysis model on the basis of existing interaction systems. The purpose is to further explore language based human-machine interaction systems and improve interaction performance. The experiment shows that the testing accuracy of the structured perceptron model reaches 95%, the recall rate reaches 81%, and the F1 value reaches 82%. The transfer dependency syntax analysis model has a data analysis speed of up to 750K/s. In simulation testing, the new robot human-machine interaction method has an accuracy of 92% compared to other existing methods, and exhibits excellent robustness and response sensitivity. In summary, research methods can provide a theoretical and practical basis for the improvement of robot interaction capabilities and the further development of human-machine collaboration.

Author 1: Shuli Wang

Author 2: Fei Long

Keywords: Human-computer interaction; speech recognition; natural language processing; lexical analysis; syntactic analysis

Paper 79: Investigating of Deep Learning-based Approaches for Anomaly Detection in IoT Surveillance Systems

Abstract: Anomaly detection plays a crucial role in ensuring the security and integrity of Internet of Things (IoT) surveillance systems. Nowadays, deep learning methods have gained significant popularity in anomaly detection because of their ability to learn and extract intricate features from complex data automatically. However, despite the advancements in deep learning-based anomaly detection, several limitations and research gaps exist. These include the need for improving the interpretability of deep learning models, addressing the challenges of limited training data, handling concept drift in evolving IoT environments, and achieving real-time performance. It is crucial to conduct a comprehensive review of existing deep learning methods to address these limitations as well as identify the most accurate and effective approaches for anomaly detection in IoT surveillance systems. This review paper presents an extensive analysis of existing deep learning methods by collecting results and performance evaluations from various studies. The collected results enable the identification and comparison of the most accurate deep-learning methods for anomaly detection. Finally, the findings of this review will contribute to the development of more efficient and reliable anomaly detection techniques for enhancing the security and effectiveness of IoT surveillance systems.

Author 1: Jianchang HUANG

Author 2: Yakun CAI

Author 3: Tingting SUN

Keywords: Internet of Things; surveillance systems; anomaly detection; deep learning; video analysis

Paper 80: A Graph-Cut Guided ROI Segmentation Algorithm with Lightweight Deep Learning Framework for Cervical Cancer Classification

Abstract: Cervical cancer classification has witnessed numerous advancements through deep learning methods; however, existing approaches often rely on multiple models for segmentation and classification, leading to heightened computational demands and prolonged training times. In this research, a lightweight deep learning framework for cervical cancer classification is presented. The framework comprises three primary components: a Graph-Cut Guided Region of Interest (ROI) segmentation algorithm, a streamlined DenseNet architecture, and a Multi-Class Logistic Regression classifier. The Graph-Cut Guided ROI segmentation algorithm is used to accurately isolate nuclei regions within multicellular Pap smear images. This is a lightweight algorithm that is able to achieve high segmentation accuracy with minimal computational overhead. The streamlined DenseNet architecture is used to efficiently extract salient features from the segmented images. This architecture is specifically designed to reduce feature redundancy and eliminate incongruous feature maps. The Multi-Class Logistic Regression classifier is used to classify the segmented images into different cell types and stages of cervical cancer. Experimental results show the proposed method is able to achieve high classification accuracy with minimal training time. The framework was trained and evaluated on a dataset of 963 Pap smear images. The proposed framework achieved a 98% cell type classification accuracy in precision, recall, and F1-score for classifying multi-cell Pap smear images. The training loss was also very low. The average training time was 21 minutes for different sets of training images, and the average testing time was 0.50 seconds for different sizes of testing images, which is much lower than the existing methods.

Author 1: Shiny T L

Author 2: Kumar Parasuraman

Keywords: Cervical cancer classification; deep learning; lightweight deep learning framework; graph-cut guided ROI segmentation algorithm; nuclei region isolation

Paper 81: Advanced Techniques for Recognizing Emotions: A Unified Approach using Facial Patterns, Speech Attributes, and Multimedia Descriptors

Abstract: The inability to efficiently store distinguishing edges, local appearance-based textured descriptions generally have limited performance in detecting facial expression analysis. The existing technology has certain drawbacks, such as the potential for poor edge-related disturbance in face photos and the reliance on present sets of characteristics that might fail to adequately represent the subtleties of emotions and thoughts in a variety of contexts. In order to overcome the difficulties associated with identifying facial expressions identification and emotion categorization, this study presents an innovative structure that combines three different information sets: a new multimedia descriptors, prosodic functions, and Local Differential Pattern (LDP). The principal driving force is the existence of noise-induced warped and weak edges in face pictures, which result in inaccurate expressions characteristic assessment. By identifying and encoding only greater edge reactions, as opposed to standard local descriptors that the LDP approach improves the endurance of face feature extraction. Robinson Compass and Kirsch Compass Masks are used for recognising edges, and the LDP formulation encodes each pixel with seven bits of information to reduce code repetition. The last category comprises Long-Term Average Spectrum (LTAS) obtained from signals related to speech, Mel-Frequency Cepstral Coefficients (MFCC), and Forming agents. Fisher Criterion is used to reduce dimensionality, and unpredictable characteristics are used in picking features. Emotion prediction is achieved by classifying two distinct circumstances using Support Vector Machine (SVM) and Decision Tree (DT) algorithms, and combining the obtained data. The research also presents a unique Visual or audio Descriptors that gives priority to key structure selections and face positioning for Audio-visual input. A concise depiction of expression is offered by the suggested Self-Similarity Distance Matrix (SSDM), which uses facial highlight points to estimate both time and space correlations. Formant frequency range, energy sources, probabilistic properties, and spectroscopic aspects define the acoustic signal. The 98% accuracy rate is attained by the emotion recognition algorithm. Major improvements upon cutting-edge techniques are shown in validation studies on the SAVEE and RML information sets, highlighting the usefulness of the suggested model in identifying and categorising emotions and facial movements in a variety of contexts. The implementation of this research is done by using Python tool.

Author 1: Kummari Ramyasree

Author 2: Chennupati Sumanth Kumar

Keywords: Local Difference Pattern (LDP); Mel-Frequency Cepstral Coefficients (MFCC); Long-Term Average Spectrum (LTAS); Self-Similarity Distance Matrix (SSDM); Support Vector Machine (SVM)

Paper 82: Research on Efficient CNN Acceleration Through Mixed Precision Quantization: A Comprehensive Methodology

Abstract: To overcome challenges associated with deploying Convolutional Neural Networks (CNNs) on edge computing devices with limited memory and computing resources, we propose a mixed-precision CNN calculation method on a Field Programmable Gate Array (FPGA). This approach involves a collaborative design encompassing both software and hardware aspects. Initially, we devised a CNN quantization method tailored for the fixed-point operation characteristics of FPGA, addressing the computational challenges posed by floating-point parameters. We introduce a bit-width strategy search algorithm that assigns bit-widths to each layer based on CNN loss variation induced by quantization. Through retraining, this strategy mitigates the degradation in CNN inference accuracy. For FPGA acceleration design, we employ a flow processing architecture with multiple Processing Elements (PEs) to support mixed-precision CNNs. Our approach incorporates a folding design method to implement shared PEs between layers, significantly reducing FPGA resource usage. Furthermore, we designed a data reading method, incorporating a register set buffer between memory and processing elements to alleviate issues related to mismatched data reading and computing speeds. Our implementation of the mixed-precision ResNet20 model on the Kintex-7 Eco R2 development board achieves an inference accuracy of 91.68% and a computing speed 4.27 times faster than the Central Processing Unit (CPU) on the CIFAR-10 dataset, with an accuracy drop of only 1.21%. Compared to a unified 16-bit FPGA accelerator design method, our proposed approach demonstrates an 89-fold increase in computing speed while maintaining similar accuracy.

Author 1: Yizhi He

Author 2: Wenlong Liu

Author 3: Muhammad Tahir

Author 4: Zhao Li

Author 5: Shaoshuang Zhang

Author 6: Hussain Bux Amur

Keywords: Convolutional Neural Networks (CNNs); edge computing technologies; Field Programmable Gate Array (FPGA) accelerator; mixed precision quantization; loss variation

Paper 83: Combining Unsupervised and Supervised Learning to Predict Poverty Households in Sakon Nakhon, Thailand

Abstract: Poverty is a problem that various government agencies are attempting to address accurately and precisely. This solution relies on data and analysis of features affecting poverty. Machine Learning is a technique to analyze and focus on poverty features encompassing five livelihood capitals: human, physical, economic, natural, and social capital to understand the household context and environment. The dataset contains 1,598 poverty households from Kut Bak district, Sakon Nakhon, Thailand. K-prototype was used to group categorical and numerical dataset into four clusters and labelled as Destitute, Extreme poor, Moderate poor, and Vulnerable non-poor. The performances of the Decision tree classifier with feature selection algorithms, including MI, ReliefF, RFE, and SFS, are compared. The best performance is SFS with F-measure, precision, and recall at 74.6%, 74.8%, and 74.7%, respectively. The result is the decision tree rules to predict the poverty level of households, enabling the establishment of guidelines for resolving household issues, and addressing broader problems within the areas.

Author 1: Sutisa Songleknok

Author 2: Suthasinee Kuptabut

Keywords: K-prototype; decision tree; feature selection; Sakon Nakhon poverty households; unsupervised learning; supervised learning

Paper 84: Network Oral English Teaching System Based on Speech Recognition Technology and Deep Neural Network

Abstract: With the development of computer technology, computer-aided instruction is being used more and more widely in the field of education. Based on speech recognition technology and deep neural network, this paper proposes an online oral English teaching system. Firstly, the speech recognition technology is introduced and its feature extraction is elaborated in detail. Then, three basic problems and three basic algorithms that need to be solved in speech recognition system using Markov model are discussed. The application of HMM technology in speech recognition system is studied, and some algorithms are optimized. The logarithmic processing of Viterbi algorithm, compared with the traditional algorithm, greatly reduces the amount of computation and solves the overflow problem in the operation process. By combining deep network with HMM, continuous speech signal modeling is realized. According to the characteristics of the DNN-HMM model, it is proposed that the model cannot model the long-term dependence of speech signals and train complex problems. Based on Kaldi, the model training comparison experiments of monophonon model, triphonon model and adding feature transformation technology are carried out to continuously improve the model performance. Finally, through simulation experiments, it is found that the recognition rate of the optimized DNN-HMM mixed model proposed in this paper is the highest, reaching 97.5%, followed by the HMM model, which is 95.4%, and the lowest recognition rate is the PNN model, which is 90.1%.

Author 1: Na He

Author 2: Weihua Liu

Keywords: Deep neural network; Markov model; voice design technology; Viterbi algorithm; oral English teaching

Paper 85: Recurrence Prediction and Risk Classification of COPD Patients Based on Machine Learning

Abstract: In response to the frequent recurrence and readmission of patients with chronic obstructive pulmonary disease, a machine learning based recurrence risk prediction and risk classification model for patients with chronic obstructive pulmonary disease is studied and constructed. Approach: This model first utilizes the optimized long short-term memory network to recognize named entities in patient electronic medical records and extract entity features. Then, XGBoost is used to predict the probability of patient relapse and readmission, and its risk is classified. Results: These results confirm that the optimized bidirectional long short-term memory network has the best performance with an accuracy of 84.36% in electronic medical record named entity recognition. The accuracy of XGBoost is the highest on both the training and testing sets, with values of 0.8827 and 0.8514, respectively. XGBoost has the best predictive ability and effectiveness. By using k-means for layering, the workload of manual evaluation was reduced by 91%, and the overall simulation accuracy of the model was as high as 97.3% and 96.4%. Conclusions: These indicate that this method can be used to balance high-risk patients between risk, cost, and resources.

Author 1: Xin Qi

Author 2: Hong Chen

Keywords: Machine learning; COPD; BiLSTM; XGBoost; k-means; recurrence; risk classification

Paper 86: Supply Chain Disturbance Management Scheduling Model Based on HPSO Algorithm

Abstract: The continuous expansion of business has led to the development of enterprises from vertical integration to horizontal integration, and the interlocking of the supply chain system, but the influence of anti-production behavior factors and the frequent occurrence of disruption events lead to difficulties in supply chain scheduling, which affects the development of enterprises. To address the above problems, the study analyzes the factors influencing counterproductive behavior based on system dynamics, constructs a supply chain disruption management scheduling model on this basis, and solves the supply chain disruption management scheduling model using Hybrid Particle Swarm Optimization algorithm. The findings indicate that the number of non-inferior solutions, uniformity of distribution of non-inferior solutions, dominance ratio of non-inferior solutions, average distance between non-inferior solutions and optimal Pareto, maximum distance, dispersion of non-inferior solutions and coverage of non-inferior solutions of the hybrid particle swarm algorithm are 12.3, 5.283, 0.264, 0.611, 4.474, 4.627, 601.300, respectively in the A condition, 601.300. The number of non-inferior solutions, uniformity of non-inferior solution distribution, dominance ratio of non-inferior solutions, average distance between non-inferior solutions and optimal Pareto, maximum distance, dispersion of non-inferior solutions and coverage of non-inferior solutions for the hybrid particle swarm algorithm under B condition are 12.3, 5.283, 0.264, 0.611, 4.474, In summary, the proposed algorithm has excellent performance and can effectively reduce the impact of interference events, thereby improving the level of supply chain interference management and scheduling, and promoting the sustainable development of this field.

Author 1: Ling Wang

Keywords: HPSO algorithm; disturbance management; supply chain; system dynamics; anti-production behavior

Paper 87: Presentation of a New Method for Intrusion Detection by using Deep Learning in Network

Abstract: Intrusion detection in cyberspace is an important field for today's research on the scope of the security of computer networks. The purpose of designing and implementing the systems of intrusion detection is to accurately categorize the virtual users, the hackers and the network intruders based on their normal or abnormal behavior. Due to the significant increase in the volume of the exchanged data in cyberspace, the identification and the reduction of inappropriate data characteristics will play a significant role in the increment of accuracy and speed of intrusion detection systems. The most advanced systems for intrusion detection are designed for the detection of an attack with the inspection of the full data of an attack. It means that a system of detection will be able to recognize the attack only after the execution of the attack on the attacked computer. In this paper, a system for end-to-end early intrusion detection is presented for the prevention of attacks on the network before these attacks cause further detriment to the system. The proposed method uses a classifier based on the network of the deep neural for the detection of an attack. The proposed network on a supervised method is trained for the exploitation of the related features by the raw data of the traffic of the network. Experimentally, the proposed approach has been evaluated on the dataset of NSL-KDD. The extensive experiments show that the presented approach performs better than the advanced approaches based on the accuracy, the rate of detection and the rate of the false positive, and also, the proposed system betters the rate of detection for the classes of the minority.

Author 1: Hui MA

Keywords: Attack; security on cyberspace; classification; intrusion detection; deep learning

Paper 88: The Application of Cognitive Decision-Making Algorithm in Cross-Border e-Commerce Digital Marketing

Abstract: Extensive global research aims to improve digital marketing profits through pricing decision-making and optimization. A dual word-of-mouth diffusion pricing model is developed for cross-border e-commerce, addressing word-of-mouth accumulation and information diffusion effects. The traditional artificial bee colony algorithm is optimized with security domain search and information diffusion profiles, enhancing global search capabilities. Performance tests reveal that word-of-mouth scale significantly influences cross-border e-commerce profits, increasing with scale coefficient, consumer conversions, and optimal profits. The proposed algorithms demonstrate high efficiency and convergence rates, surpassing common iterations and benefits in the clothing pricing problem. The comprehensive imitation effect is -0.14, and the word-of-mouth scale effect is 1.34. Pre-sale and sales prices for clothing are set at 347.49 and 641.393, respectively. Similarly, in pricing cross-border e-commerce electronic products, the algorithm achieves optimal profits after 230 iterations, surpassing other algorithms. Overall, the proposed model exhibits superior computational performance in cross-border e-commerce pricing decision-making compared to conventional approaches.

Author 1: Xuehui Wang

Keywords: Cross-border e-commerce; decision-making; pricing issues; optimization algorithms; ABO

Paper 89: Planning and Expansion of the Transmission Network in the Presence of Wind Power Plants

Abstract: The proliferation of renewable energy sources, particularly wind farms, is rapidly gaining momentum owing to their numerous benefits. Consequently, it is imperative to account for the impact of wind farms on transmission expansion planning (TEP), which is a crucial aspect of power system planning. This article presents a multi-objective optimization model that utilizes DC load flow to address the TEP challenge while also incorporating wind farm uncertainties into the model. The present study aims to optimize the expansion and planning of the TEP in the power system by considering investment and maintenance costs as objective functions. To achieve this, a multi-objective approach utilizing the shuffled frog leaping algorithm (SFLA) is proposed and implemented. The proposed objectives are simulated on the RTS-IEEE 24-bus test network. The results obtained from the proposed algorithm are compared with those of the Genetic Algorithm (GA) to assess and validate the proposed approach.

Author 1: Hui Sun

Keywords: Wind farms; Transmission Expansion Planning (TEP); multi-objective optimization model; Shuffled Frog Leaping Algorithm (SFLA)

Paper 90: LPDA: Cross-Project Software Defect Prediction Approach via Locality Preserving and Distribution Alignment

Abstract: Cross-Project Defect Prediction (CPDP) based on domain adaptation aims to achieve defect prediction tasks in an unlabeled target software project by borrowing the defect knowledge extracted from well-annotated source software projects. Most existing CPDP approaches enhance transferability between projects but struggle with misalignments due to limited exploration of class-specific features and inability to preserve original local relationships in transformed features. In order to tackle these challenges, The article introduces a novel Cross-Project Defect Prediction (CPDP) approach called Local Preserving and Distribution Alignment (LPDA). This approach addresses the challenge of misalignments in CPDP due to limited exploration of discriminative feature representations and the failure to preserve original local relationship consistency. LPDA combines transferability and discriminability for CPDP tasks. It uses locality-preserving projection to maintain module consistency and distribution alignment, which includes transferable and discriminant distribution alignment. The former narrows the distributions of both source and target projects, while the latter increases the discrepancy between different classes across projects. The effectiveness of LPDA was tested through 118 cross-project prediction tasks involving 22 software projects from four distinct repositories. The results showed that LPDA outperforms baseline CPDP methods by efficiently learning representations that integrate transferability and discriminability while preserving local geometry to optimize distances within and between categories.

Author 1: Jin Xian

Author 2: Jinglei Li

Author 3: Quanyi Zou

Author 4: Yunting Xian

Keywords: Cross Project Defect Prediction; discriminative distribution alignment; local preserving; domain adaption

Paper 91: Exploiting Deepfakes by Analyzing Temporal Feature Inconsistency

Abstract: In recent years, the rapid advancement of image generation technology has facilitated the creation of counterfeit images and videos, posing significant challenges for content authenticity verification. Malefactors can easily extract videos from social networks and generate their own deceptive renditions using state-of-the-art techniques. The latest Deepfake face forgery videos have reached an unprecedented level of sophistication, making it exceptionally difficult to discern signs of manipulation. While several methods have been proposed for detecting fraudulent media, they often target specific aspects, and as new attack methods emerge, these approaches tend to become obsolete. This paper presents a novel detection approach that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory Networks (LSTM). Initially, CNN is employed to extract image features from each frame of the input facial video, capturing subtle alterations and irregularities in manipulated content. Subsequently, the extracted feature sequence is used to train the LSTM network, mimicking the temporal consistency of human visual perception and enhancing the effectiveness of counterfeit video detection. To validate this methodology, a comprehensive evaluation is conducted using the FaceForensic++ dataset, affirming its proficiency in identifying Deepfake counter-feit videos.

Author 1: Junlin Gu

Author 2: Yihan Xu

Author 3: Juan Sun

Author 4: Weiwei Liu

Keywords: Face forgery detection; Convolutional Neural Net-work; Long Short-Term Memory Network; time consistency

Paper 92: The PSR-Transformer Nexus: A Deep Dive into Stock Time Series Forecasting

Abstract: Accurate stock market forecasting has remained an elusive endeavor due to the inherent complexity of financial systems dynamics. While deep neural networks have shown initial promise, robustness concerns around long-term dependencies persist. This research pioneers a synergistic fusion of nonlinear time series analysis and algorithmic advances in representation learning to enhance predictive modeling. Phase space reconstruction provides a principled way to reconstruct multidimensional phase spaces from single variable measurements, elucidating dynamical evolution. Transformer networks with self-attention have recently propelled state-of-the-art results in sequence modeling tasks. This paper introduces PSR-Transformer Networks specifically tailored for stock forecasting by feeding PSR interpreted constructs to transformer encoders. Extensive empirical evaluation on 20 years of historical equities data demonstrates significant accuracy improvements along with enhanced robustness against LSTM, CNN-LSTM and Transformer models. The proposed interdisciplinary fusion establishes new performance benchmarks on modeling financial time series, validating synergies between domain-specific reconstruction and cutting-edge deep learning.

Author 1: Nguyen Ngoc Phien

Author 2: Jan Platos

Keywords: Stock market forecasting; deep learning; chaos theory; phase space reconstruction; transformer neural networks; time series analysis

Paper 93: Efficient IoT Security: Weighted Voting for BASHLITE and Mirai Attack Detection

Abstract: The increasing number of devices in the Internet of Things (IoT) has exposed various vulnerabilities, such as BASHLITE and Mirai attacks, making it easier for cyber threats to emerge. Due to these vulnerabilities, developing innovative detection and mitigation strategies is essential. Our proposed solution is an ensemble-based weighted voting model that combines different classifiers, including Random Forest, eXtreme Gradient Boosting (XGBoost), Gradient Boosting, K-nearest neighbor (KNN), Multilayer Perceptron (MLP), and Adaptive Boosting (AdaBoost), using artificial intelligence and machine learning. We evaluated our model on the N-BaIoT dataset, a benchmark in this domain. Our results show that the weighted voting approach has exceptional accuracy, precision, recall, and F1-Score. This highlights the effectiveness of our model in classifying various attack instances within the IoT security context. Our approach performs better than other state-of-the-art methods, achieving a remarkable accuracy of 99.9955% in detecting and preventing BASHLITE and Mirai cyber-attacks on IoT devices.

Author 1: Marwan Abu-Zanona

Keywords: Internet of Things; IoT security; BASHLITE; Mirai attacks; ensemble learning

Paper 94: Sentiment Analysis on Banking Feedback and News Data using Synonyms and Antonyms

Abstract: Sentiment analysis is crucial for deciphering customers’ enthusiasm, frustration, and the market mood within the banking sector. This importance arises from financial data’s specialized and sensitive nature, enabling a deeper understanding of customer sentiments. In today’s digital and social marketing landscape within the banking and financial sector, sentiment analysis is significant in shaping customer insights, product development, brand reputation management, risk management, customer service improvement, fraud detection, market research, compliance regulations, etc. This paper introduces a novel approach to sentiment analysis in the banking sector, emphasizing integrating diverse text features to enable dynamic analysis. This proposed approach aims to assess the sentiment score of distinct words used within a document and classify them as positive, negative, or neutral. After rephrasing sentences using synonyms and antonyms of unique words, the system calculates sentence similarity using a distance control mechanism. Then, the system updates the dataset with the positive, negative, and neutral labels. Ultimately, the ELECTRA model utilizes the self-trained sentiment-scored data dictionary, and the newly created dataset is processed using the SoftMax activation function in combination with a customized ADAM optimizer. The approach’s effectiveness is confirmed through the analysis of post-bank customer feedback and the phrase bank dataset, yielding accuracy scores of 92.15%and 93.47%, respectively. This study stands out due to its unique approach, which centers on evaluating customer satisfaction and market sentiment by utilizing sentiment scores of words and assessing sentence similarities.

Author 1: Aniruddha Mohanty

Author 2: Ravindranath C. Cherukuri

Keywords: ELECTRA; Synonyms and antonyms; sentiment analysis; datasets; sentiment score; control distance

Paper 95: Iterative Learning Control for High Relative Degree Discrete-Time Systems with Random Initial Shifts

Abstract: In this paper, an iterative learning control (ILC) strategy under compression mapping framework is presented for high relative degree discrete-time systems with random initial shifts. Firstly, utilizing the high relative degree of the system and difference term, a control law is designed and a p-order non-homogeneous linear difference equation is established. The appropriate control gain is selected according to the charac-teristics of solution of the difference equation and the initial shifts, so as to ensure that the high relative degree discrete-time system can reach a steady-state deviation output at a fixed time. Subsequently, a PD-type control law is employed to correct the fixed deviation of the system. Theoretical analysis indicates that this ILC strategy can ensure that the high relative degree systems achieve accurate tracking after the predefined time. Finally, the simulation experiments are conducted on a linear discrete-time Multiple-Input Multiple-Output(MIMO) system with relative degree 1 and a Multiple-Input Single-Output(MISO) system with relative degree 2, respectively, and the results verify the effectiveness of the algorithm.

Author 1: Dongjie Chen

Author 2: Tiantian Lu

Author 3: Zhenjie Yin

Keywords: Relative degree; iterative learning control; random initial shifts; difference equation; discrete-time system

Paper 96: A Memetic Algorithm to Solve the Two-Echelon Collaborative Multi-Centre Multi-Periodic Vehicle Routing Problem with Specific Constraints

Abstract: The collaboration between distribution companies is gaining a great interest in the last years due to the benefit provided to reduce the cost of deliveries. In this work we study the centralized two-echelon collaborative multi-center multi-periodic vehicle routing problem with a specific constraints. In which each distribution center conserves its VIP customers, and each partner keep their delivery scheduling unchangeable. The problem is modelled as a MILP, and to solve it a hybrid algorithm is proposed. This algorithm combines a multi-population memetic algorithm (MPMA) and a variable neighbourhood search algorithm that integrates a tabu search list (VNS-T). The results obtained are compared with those obtained by CPLEX solver and the best known solution of the multi-depot vehicle routing problem (MDVRP).

Author 1: Camelia Snoussi

Author 2: Abdellah El Fallahi

Author 3: Sarir Hicham

Keywords: Collaborative vehicle routing problem; two-echelon networks; memetic algorithm; Variable neighbourhood serach

Paper 97: Speech Recognition Models for Holy Quran Recitation Based on Modern Approaches and Tajweed Rules: A Comprehensive Overview

Abstract: Speech is considered the most natural way to communicate with people. The purpose of speech recognition technology is to allow machines to recognize and understand human speech, enabling them to take action based on the spoken words. Speech recognition is especially useful in educational fields, as it can provide powerful automatic correction for language learning purposes. In the context of learning the Quran, it is essential for every Muslim to recite it correctly. Traditionally, this involves an expert gari who listens to the student’s recitation, identifies any mistakes, and provides appropriate corrections. While effective, this method is time-consuming. To address this challenge, apps that help students fix their recitation of the Holy Quran are becoming increasingly popular. However, these apps require a robust and error-free speech recognition model. While recent advancements in speech recognition have produced highly accurate results for written and spoken Arabic and non-Arabic speech recognition, the field of Holy Quran speech recognition is still in its early stages. Therefore, this paper aims to provide a comprehensive literature review of the existing research in the field of Holy Quran speech recognition. Its goal is to identify the limitations of current works, determine future research directions, and highlight important research in the fields of spoken and written languages.

Author 1: Sumayya Al-Fadhli

Author 2: Hajar Al-Harbi

Author 3: Asma Cherif

Keywords: Speech recognition; acoustic models; language model; neural network; deep learning; quran recitation

Paper 98: Improving the Classification of Airplane Accidents Severity using Feature Selection, Extraction and Machine Learning Models

Abstract: Airplane mode of transportation is statistically the most secure means of travel. This is due to the fact that flights require several conditions and precautions because aviation accidents are most of the time fatal and have disastrous consequences. For this purpose, in this paper, the mean goal is to study the different levels of fatality of airplane accidents using machine learning models. The study rely on airplane accident severity dataset to implement three machine learning models: KNN, Decision Tree and Random Forest. This study began with implementing two features selection and extraction methods, PCA and RFE in order to reduce dataset dimensionality and complexity of models and reduce training time by implementing machine learning models on dataset and measuring their performance. Results show that KNN and Decision Tree demonstrates high levels of performances by achieving 100% of accuracy and f1-score metrics; while Random Forest achieves its best performances after application of PCA when it reaches an accuracy equal to 97.83% and f1-score equal to 97.82%.

Author 1: Rachid KAIDI

Author 2: Mohammed AL ACHHAB

Author 3: Mohamed LAZAAR

Author 4: Hicham OMARA

Keywords: Airplane accident; severity; flights safety; machine learning; KNN; Random Forest (RF); Decision Tree (DT)

Paper 99: Enhancing Safety and Multifaceted Preferences to Optimise Cycling Routes for Cyclist-Centric Urban Mobility

Abstract: In order to optimise bicycle routes across a variety of multiple parameters, including safety, efficiency and subtle rider preferences, this work explores the difficult domain of the Bike Routing Problem (BRP) using a sophisticated Simulated Annealing approach. In this innovative structure, a wide range of limitations and inclinations are combined and carefully calibrated to create routes that skillfully meet the varied and changing needs of cyclists. Extensive testing on a dataset representing a range of rider preferences demonstrates the effectiveness of this novel approach, resulting in significant improvements in route selection. This research is a significant resource for urban planners and politicians. Its data-driven solutions and strategic recommendations will help them strengthen bicycle infrastructure, even beyond its immediate applicability in resolving the BRP.

Author 1: Mohammed Alatiyyah

Keywords: Bike routing; dynamic vehicle routing inventory routing; approximate dynamic programming

Paper 100: Recognition and Translation of Ancient South Arabian Musnad Inscriptions

Abstract: Inscriptions play an important role in preserving historical information. As such, conservation of these inscriptions provides valuable insights into the history and cultural heritage of the region. Musnad inscriptions are considered one of the earliest forms of writing from the Arabian Peninsula, preceding the modern Arabic font; however, most Musnad inscriptions remain unread and untranslated, signifying a substantial loss of historical information. In response, this paper represents a significant contribution to the field by proposing a successful approach to interpreting Musnad inscriptions. To do so, a dataset was prepared from the Saudi Arabian Ministry of Culture and subjected to preprocessing for optimal recognition, a step that entailed several experiments to enhance image quality and preparedness for recognition. The dataset was then trained and tested with 29 classes using three different convolutional neural network (CNN) architectures: Visual Geometry Group 16 (VGG16), Residual Network 50 (ResNet50) and MobileNetV2. Thereafter, the performance of each architecture was evaluated based on its accuracy in recognising Musnad inscriptions. The results demonstrate that VGG16 achieved the highest accuracy of 93.81%, followed by ResNet50 at 89.39% and MobileNetV2 at 80.02%.

Author 1: Afnan Altalhi

Author 2: Atheer Alwethinani

Author 3: Bashaer Alghamdi

Author 4: Jumanah Mutahhar

Author 5: Wojood Almatrafi

Author 6: Seereen Noorwali

Keywords: Musnad inscriptions; text recognition; deep learning; VGG16; ResNet-50; MobileNetV2

Paper 101: Learnable Local Similarity for Face Forgery Detection and Localization

Abstract: The emergence of many face forgery technologies has led to the widespread of forgery faces on the Internet, causing a series of serious social impacts, thus face forgery detection technology has attracted increasing attention. While many face forgery detection algorithms have demonstrated impressive performance against known manipulation methods, their efficacy tends to diminish severely when applied to unknown forgeries. Previous research commonly viewed face forgery detection as a binary classification problem, disregarding the crucial distinction between real and forged faces, thereby limiting the generalizability of detection algorithms. To overcome this issue, this paper proposes a novel face forgery detection method that utilizes a trainable metric to learn local similarity between local features of facial images, achieving a more generalized detection result. What’s more, it incorporate cross-level features to accurately locate forgery regions. After conducting extensive experiments on FaceForensics++, Celeb-DF-v2, and DFD, which demonstrate that the effectiveness of the proposed method is comparable to state-of-the-art detection algorithms.

Author 1: Lingyun Leng

Author 2: Jianwei Fei

Author 3: Yunshu Dai

Keywords: Face forgery detection; local similarity; forgery localization; generalized detection

Paper 102: CLFM: Contrastive Learning and Filter-attention Mechanism for Joint Relation Extraction

Abstract: Relation extraction is a fundamental task in natural language processing, which involves extracting structured information from textual data. Despite the success of joint methods in recent years, most of them still have the propagation of cascade errors. Specifically, the error in former step will be accumulated into the final combined triples. Meanwhile, these methods also encounter another challenges related to insufficient interaction between subtasks. To alleviate these issues, this paper proposes a novel joint relation extraction model that integrates a contrastive learning approach and a filter-attention mechanism. The proposed model incorporates a potential relation decoder that utilizes contrastive learning to reduce error propagation and enhance the accuracy of relation classification, particularly in scenarios involving multiple relationships. It also includes a relation-specific sequence tagging decoder that employs a filter-attention mechanism to highlight more informative features, alongside an auxiliary matrix that amalgamates information related to entity pairs. Extensive experiments are conducted on two public datasets and the results demonstrate that this approach outperforms other models with the same structure in recall and F1. Moreover, experiments show that both the contrastive learning strategy and the proposed filter-attention mechanism work well.

Author 1: Zhiyuan Wang

Author 2: Chuyuan Wei

Author 3: Jinzhe Li

Author 4: Lei Zhang

Author 5: Cheng Lv

Keywords: Natural language processing; relation extraction; attention mechanism; contrastive learning; multi-task learning

Paper 103: Enhancing Airborne Disease Prediction: Integrating Deep Infomax and Self-Organizing Maps for Risk Factor Identification

Abstract: Asthma poses a significant global public health concern, particularly in urban centers where environmental pollutants and variable weather patterns contribute to heightened prevalence and symptom exacerbation. The Deonar dumping ground, one of Mumbai’s largest landfills, releases a complex mix of particulate matter and hazardous gases, posing a serious threat to local respiratory health. Despite the urgency for comprehensive research integrating patient-specific data with localized weather and air quality metrics, such studies remain limited. This study addresses the critical research gap by investigating asthma risk factors near the Deonar dumping ground. Integrating detailed patient records with precise local weather and air quality measurements, our research aims to unravel the intricate relationship between environmental exposure and respiratory health outcomes. The findings provide crucial insights into the specific risk factors influencing asthma incidence and severity in this region, informing the development of targeted interventions and mitigation strategies. Employing a novel ensemble Deep Info Max - Self-Organizing Map (DIM-SOM) technique, our study compares its performance with various clustering algorithms, including SOM, K-Means, Bisecting K-Means, DBSCAN, and others. The novel ensemble DIM-SOM demonstrated superior performance, achieving a significantly higher Silhouette Score of 0.9234, a lower Davies-Bouldin Score of 0.1276, and a more favorable Calinski-Harabasz Score of 389723.6225 compared to other algorithms. These findings underscore the efficacy of the novel ensemble DIM-SOM approach in generating dense, well-separated, and meaningful clusters, emphasizing its potential to enhance clustering performance compared to traditional algo-rithms. The study further emphasizes the need for proactive mitigation measures and tailored healthcare interventions based on the identified environmental risk factors.

Author 1: Bhakti S. Pimpale

Author 2: Anala A. Pandit

Keywords: Asthma; deepinfomax; self organizing map; risk factors; air pollution

Paper 104: Enhancing Assamese Word Recognition for CBIR: A Comparative Study of Ensemble Methods and Feature Extraction Techniques

Abstract: This study conducts a thorough assessment of ensemble machine learning methods, specifically focusing on the identification of Assamese words. This task is crucial for improving Content-Based Image Retrieval systems and safeguarding the digital heritage of Assamese culture. We analyze the efficacy of different algorithms, such as CatBoost, XGBoost, Gradient Boosting, Random Forest, Bagging, AdaBoost, Stacking, and Histogram-Based Gradient Boosting, by thoroughly examining their performance in terms of accuracy, precision, recall, Kappa, F1-score, Matthews Correlation Coefficient, and AUC. The Cat-Boost algorithm stands out as the top performer, achieving an accuracy rate of 97.7%, precision rate of 95%, and recall rate of 96%. XGBoost is also acknowledged for its substantial effectiveness. This comparative analysis emphasizes CatBoost’s superiority in terms of precision and recall. Additionally, it underscores the strong ability of ensemble classifiers to enhance assistive technologies, promote social inclusivity, and seamlessly integrate the Assamese language into technological applications.

Author 1: Naiwrita Borah

Author 2: Udayan Baruah

Author 3: Barnali Dey

Author 4: Merin Thomas

Author 5: Sunanda Das

Author 6: Moumi Pandit

Author 7: Bijoyeta Roy

Author 8: Amrita Biswas

Keywords: Assamese literary works; automatic word recognition; comparative analysis; feature-based approaches; intelligent assistive technology; machine learning; word image analysis

Paper 105: A Hybrid Deep Learning Framework for Efficient Sentiment Analysis

Abstract: In the era of Microblogging and the rapid growth of online platforms, an exponential rise is shown in the volume of data generated by internet users across various domains. Additionally, the creation of digital or textual data is expanding significantly. This is because consumers respond to comments made on social media platforms regarding events or products based on their personal experiences. Sentiment analysis is usually used to accomplish this kind of classification on a large scale. It is described as the process of going through all user reviews and comments that are discovered in product reviews, events, or similar sources in order to look for unstructured text comments. Our study examines how deep learning models like LSTM, GRU, CNN, and hybrid models (LSTM+CNN, LSTM+GRU, GRU+CNN) capture complex sentiment patterns in text data. Additionally, we study integrating BOW and TF-IDF as complementing features to improve model predictive power. CNN with RNNs consistently improves outcomes, demonstrating the synergy between convolutional and recurrent neural network architectures in recognizing nuanced emotion subtleties.In addition, TF-IDF typically outperforms BOW in enhancing deep learning model sentiment analysis accuracy.

Author 1: Asish Karthikeya Gogineni

Author 2: S Kiran Sai Reddy

Author 3: Harika Kakarala

Author 4: Yaswanth Chowdary Gavini

Author 5: M Pavana Venkat

Author 6: Koduru Hajarathaiah

Author 7: Murali Krishna Enduri

Keywords: Sentiment analysis; LSTM; GRU; Convolutional Neural Networks (CNNs); BOW; TF-IDF

Paper 106: Predictive Modeling of Landslide Susceptibility in Soft Soil Canal Regions: A Focus on Early Warning Systems

Abstract: The Mekong Delta (MD) has suffered significant losses in land resources, economic damage, and human and property casualties due to recent landslides. An early warning system for landslides is a valuable tool for identifying the effectiveness and timely detection of changes in the soil to promptly determine solutions and minimize damage caused by landslides in an area. In this study, we apply a machine learning approach based on the Long Short-Term Memory (LSTM) algorithm to experiment with early warning of landslide events on soft soil in the MD. Horizontal pressure, the change in inclination angles of the sensor pile due to the soil mass sliding in both the x and y directions, and the warning levels are determined based on the deformation and displacement of the soil along the riverbank, considered candidate factors for inputs in the model. Data from the established sensor system is used to train the model, creating a training and testing dataset of 374,415 samples. The accuracy of the detection and classification threshold of the system is proposed to be measured using the average F1 score derived from precision and recall values. The optimal prediction results are gleaned from an observational window of 4 minutes and 30 seconds to project roughly 2 hours into the future. The validation process resulted in recall, precision, and F1-score stands at 0.8232 with a remarkably low standard deviation of about 1%. The successful application of this research can help identify abnormal events leading to riverbank landslides due to loading, thereby creating conditions for developing a reliable information system to provide managers with the ability to suggest timely solutions to protect the lives, property of residents and infrastructures.

Author 1: Dang Tram Anh

Author 2: Luong Vinh Quoc Danh

Author 3: Chi-Ngon Nguyen

Keywords: Landslide early warning; soft soil; Mekong Delta; long short-term memory

Paper 107: Transformative Learning Through Augmented Reality Empowered by Machine Learning for Primary School Pupils: A Real-Time Data Analysis

Abstract: Academic performance and student engagement are constant challenges in the field of modern education. When it comes to engaging students, traditional teaching methods frequently fall short, so creative solutions are needed. The Transformative potential of Augmented Reality (AR) technology as a cutting-edge teaching strategy is examined in this study. AR presents a dynamic, immersive learning environment that has the potential to completely transform conventional class-rooms. By incorporating AR into the curriculum, our research transforms pedagogical paradigms, closes the engagement gap, and raises academic performance through an adaptive learning system. The study reveals the complex dynamics of AR-enhanced education through thorough analysis, powerful visualizations, and significant ANOVA results (p-value=0.03). It challenges accepted educational theories and provides insights into the complex effects on learning outcomes and student engagement. This study highlights the significance of AR in educational settings and promotes its incorporation as a transformative instrument that can establish dynamic and captivating learning environments, encourage critical thinking, creativity, and early field exploration through Artificial Intelligence (AI), and ultimately mould future leaders who can succeed.

Author 1: Abinaya M

Author 2: Vadivu G

Keywords: Artificial intelligence; augmented reality; adaptive learning; machine learning; transformative learning

Paper 108: Indonesian Twitter Emotion Recognition Model using Feature Engineering

Abstract: Twitter is a social media platform that has a large amount of unstructured natural language text. The content of Twitter can be utilized to capture human behavior via emphasized emotions located in tweets. In their tweets, people commonly express emotions to show their feelings. Hence, it is crucial to recognize the text’s underlined emotions to understand the message’s meaning. Feature engineering is the process of improving raw data into often overlooked features. This research explores feature engineering techniques to find the best features for building an emotion recognition model on the Indonesian Twitter dataset. Two different text data representations were used, namely, TF-IDF and word embedding. This research proposed 12 feature engineering configurations in TF-IDF by combining data stemming, data augmentation, and machine learning classifiers. Moreover, this research proposed 27 feature engineering configurations in word embedding by combining three-word embedding models, three pooling techniques, and three machine-learning classifiers. In total, there are 39 feature engineering combinations. The configuration with the best F1 score is TF-IDF with logistic regression, stemmed dataset, and augmented dataset. The model achieved 65.27% accuracy and 66.09% F1 score. The detailed characteristics from the top seven models in TF-IDF also follow the same feature engineering configuration. Lastly, this work improves performance from the previous research by 1.44% and 2.01% on the word2vec and fastText approaches, respectively.

Author 1: Rhio Sutoyo

Author 2: Harco Leslie Hendric Spits Warnars

Author 3: Sani Muhamad Isa

Author 4: Widodo Budiharto

Keywords: Text classification; feature engineering; emotion recognition; Indonesian tweet; natural language processing

Paper 109: Robust Extreme Learning Machine with Exponential Squared Loss via DC Programming

Abstract: Extreme learning machines (ELM) have recently attracted considerable attention because of its fast learning rate, simple model structure, and good generalization ability. However, classical ELM with least squares loss function is prone to overfitting and lack robustness in dealing with datasets containing noise and outliers in the real world. In this paper, inspired by the maximum correntropy criterion, an exponential squared loss function is introduced, which is nonconvex and insensitive to noise and outliers. A robust ELM with exponential squared loss (RESELM) is presented to overcome the overfitting problem. The proposed model with nonconvexity is difficult to be directly optimized. Considering the superior performance of difference of convex functions (DC) programming in solving nonconvex problems, this paper optimizes the model by expressing the objective function as a DC function and employing DC algorithm (DCA). To examine the effectiveness of the proposed algorithm in noisy environment, different levels of outliers are added to the training samples in the experiments. Experimental results on benchmark data sets with different outliers levels illustrate that the proposed RESELM achieves significant advantages in generalization performance and robustness, especially in higher outliers levels.

Author 1: Kuaini Wang

Author 2: Xiaoxue Wang

Author 3: Weicheng Zhan

Author 4: Mingming Wang

Author 5: Jinde Cao

Keywords: Extreme learning machine; exponential squared loss; DC programming; DCA; robust regression

Paper 110: Quality of Data (QoD) in Internet of Things (IOT): An Overview, State-of-the-Art, Taxonomy and Future Directions

Abstract: The Internet of Things (IoT) data is the main component for finding the basis that allows decisions to be made intelligently and enables other services to be explored and used. Data originates from smart things that have the capabilities to connect and share data enormously with other things in the IoT ecosystem. However, the level of intelligence obtained and the type of services provided, all depend on whether the data is trusted or not. High-quality data is the most trusted;, it can be used to extract meaningful insights from an event and can also be used to provide good services to humans. Therefore, decisions based on high-quality and trusted data could be good, whereas those based on low-quality or untrusted data are not only bad but could also have severe consequences. The term Quality of Data (QoD) is used to represent data trustworthiness and is used throughout this paper. To the best of our knowledge, this work is the first to coin the term QoD. The problems that hinder QoD are identified and discussed. One if it is an outlier, it is a major feature of the data that degrades its overall quality. Several machine-learning techniques that detect outliers have been studied and presented, with few data-cleaning techniques. This paper aims to present the elements necessary to ensure QoD by presenting the overview of the IoT state-of-the-art. Then, data quality, data in IoT, and outliers are studied, and some quality assurance techniques that maintain data quality is presented. A comprehensive taxonomy is shown to provide state-of-the-art data in IoT. Open issues and future directions were suggested at the end of the paper.

Author 1: Jameel Shehu Yalli

Author 2: Mohd Hilmi Hasan

Author 3: Nazleeni Samiha Haron

Author 4: Mujeeb Ur Rehman Shaikh

Author 5: Nafeesa Yousuf Murad

Author 6: Abdullahi Lawal Bako

Keywords: Quality of Data (QoD); Internet of Things (IoT); RFID; WSN; Taxonomy; trustworthiness; outlier; anomaly; confusion matrix; QoD assurance technique