IJACSA Volume 15 Issue 2 - thesai.org

IJACSA Volume 15 Issue 2

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: Deep Neural Network-based Methods for Brain Image De-noising: A Short Comparison

Abstract: Various types of noise may affect the visual quality of images during capturing and transmitting procedures. Finding a proper technique to remove the possible noise and improve both quantitative and qualitative results is always considered as one of the most important and challenging pre-processing tasks in image and signal processing. In this paper, we made a short comparison between two well-known approaches called thresholding neural network (TNN) and deep neural network (DNN) based methods for image de-noising. De-noising results of TNNs, Dn-CNNs, Flashlight CNN (FLCNN) and Diamond de-noising networks (DmDN) have been compared with each other. In this regard, several experiments have been performed in terms of Peak Signal to Noise Ratio (PSNR) to validate the performance analysis of various de-noising methods. The analysis indicates that DmDNs perform better than other learning-based algorithms for de-noising brain MR images. DmDN achieved a PSNR value of 29.85 dB, 30.74 dB, 29.15 dB, and 29.45 dB for de-noising MR image 1, MR image 2, MR image 3 and MR Image 4, respectively for a standard deviation of 15.

Author 1: Keyan Rahimi

Author 2: Noorbakhsh Amiri Golilarz

Keywords: CNN; Deep neural network; de-noising; MR image; PSNR

Paper 2: Assessing and Mitigating Network Vulnerabilities in Philips Hue and Nest Protect Smart Home Devices

Abstract: The Internet of Things (IoT) has gained momentum across various sectors, particularly in the consumer market with the adoption of smart devices. IoT extends internet connectivity to physical devices, enabling control via smartphones, environmental sensing, and updates. However, smart home devices are susceptible to cyberattacks due to vulnerabilities, lack of monitoring, and built-in security. They can also participate in botnets, leading to large-scale attacks. Vulnerabilities in these devices may exist at the sensing, network, or application layers, impacting data confidentiality, integrity, and service availability. This research aims to identify network-layer vulnerabilities affecting the 'Availability' of Philips Hue and Nest Protect. By establishing a test environment, the baseline behavior of these devices is examined, followed by scans for open ports and services to detect network-based threats. Volumetric flood attacks are then conducted to assess susceptibility, and findings are shared to define the devices' default security posture. The research also addresses security issues related to home routers and aims to reduce the attack surface of smart home devices through isolation and network-level protection. This involves deploying a Firewall to isolate smart devices from non-IoT devices and prevent intrusions.

Author 1: Arvind Sredhar

Author 2: Adil Khan

Author 3: Abdul Rehman Gilal

Author 4: Aeshah Alsughayyir

Author 5: Abdullah Alshanqiti

Author 6: Bandeh Ali Talpur

Keywords: Internet of Things (IoT); Smart Home Devices (SHDs); network vulnerability assessment; Philips Hue; Nest Protect

Paper 3: Texture and Color Descriptor Features-based Vacant Parking Space Detection using K-Nearest Neighbors

Abstract: The importance of the detection of vacant parking spaces is increasing gradually. A system capable of detecting vacant parking spaces in real-time can play an important role in saving valuable time for motorists, decreasing traffic jams, and reducing air pollution. Vision-based parking space detection methods are advantageous in terms of installation and maintenance as existing security cameras in a parking area can be used without the requirement of additional hardware and the detection program can be run on a local or a remote server. One major problem of the vision-based detection method in this context is making the model generalized for detection in various weather conditions. This research proposes a hybrid method to detect vacant parking spaces that use texture and color descriptors. A weighted KNN is used for the classification of parking spaces. The proposed method experimented on PKLot, a large benchmarking dataset that contains images of three parking areas in three weather conditions. The proposed model achieves an accuracy of 99.47% on average while training with 10-fold cross-validation and achieves an accuracy of 99.41% accuracy on average while testing with unseen data. The model shows robustness and better performance in terms of accuracy and processing speed. Several comparisons are also done to show how well it performs with methods found in previous research.

Author 1: A F M Saifuddin Saif

Author 2: Zainal Rasyid Mahayuddin

Keywords: Texture; color descriptor; k-nearest neighbors; computer vision; image processing

Paper 4: SkySculptor: Intuitive Drone Control Through Ground-Integrated Radar and Foot Gestures in Smart Indoor Environments

Abstract: SkySculptor is a software application designed to optimize drone control in smart indoor environments. The primary focus is on using gesture input for drone control, particularly investigating mid-air free-foot interactions detected by radar sensing. This software application simplifies the process of controlling drones in smart indoor environments. Additionally, outcomes of utilizing a 15-antenna ultra-wideband 3D radar are presented, establishing a dictionary of six directional swipe gestures for controlling drone functions. Based on the findings of this research article, guidelines for the future development of software applications for drone control in intelligent indoor environments are proposed.

Author 1: Alexandru-Ionut Siean

Keywords: Drone control; gesture input; ultra-wideband radar; software application; smart environments

Paper 5: Classifying Motorcycle Rider Helmet on a Low Light Video Scene using Deep Learning

Abstract: For safety in transportation, it is important to always monitor the use of proper motorcycle helmet, especially at night. One way to enforce transportation rules and regulations in wearing proper motorcycle helmet is to use computer vision technology. This study focusses on classifying motorcycle rider helmet at low light video conditions, like at dusk and at night, using YOLOv5 and YOLOv7 with Deep SORT. In these deep learning methods, the study tunes and optimizes hyperparameters to attain high accuracy in classifying motorcycle rider helmet at this challenging environment. To accomplish this objective, a vast and diverse dataset was employed, containing classes such as riders, different types of helmets (valid and invalid), and instances of riders not wearing helmets at all in Metro Manila, Philippines. The results show that Hyperparameter 3 consistently outperformed other settings in terms of precision (95.6%), recall (91.2%), and mean average precision (mAP) scores across multiple scales and time frames with 95.1% on [email protected] and 76.3% on [email protected], owing to greater epochs, quicker learning rates, and lower batch sizes.

Author 1: John Paul Q. Tomas

Author 2: Bonifacio T. Doma

Keywords: Artificial intelligence; computer vision; computer vision problems; object detection; YOLOv5; YOLOv7; Deep SORT; deep learning

Paper 6: China's Science and Technology Finance and Economic Corridor Development: A Coupling Relationship Analysis

Abstract: This study aims to explore the coupling relationship between science and technology finance and economic corridor development in my country based on the life cycle theory of industrial clusters. By analyzing the interaction between science and technology finance and the development of economic corridors, our degree of correlation and influence mechanism at different stages are revealed. The main findings of this study include that at different stages of the life cycle of industrial clusters, there are differences in the effect of science and technology finance on the development of economic corridors, showing a gradually strengthening or decreasing trend; there is a strong positive relationship between science and technology finance and the development of economic corridors. Coupling relationship, the development of science and technology finance promotes the construction and development of economic corridors, and vice versa; the coupling relationship between science and technology finance and the development of economic corridors has important policy implications and provides useful information for government departments, business managers and scientific research institutions Reference and guidance. These findings are of great value to the formulation and implementation of science and technology finance policies, the planning and construction of economic corridors, and the cultivation and development of industrial clusters. Government departments can adjust science, technology, finance and economic corridor development policies based on the characteristics of the coupling relationship to promote their coordinated development and virtuous cycle. Based on the research results, business managers and scientific research institutions can optimize resource allocation, enhance innovation capabilities, and achieve sustainable development.

Author 1: Rui Tian

Author 2: Birong Xu

Keywords: Technology finance; regional economic development; industrial clusters; life cycle theory

Paper 7: Correlation Analysis Between Student Psychological State and Grades Based on Data Mining Algorithms

Abstract: As society has evolved and educational reform has become more profound, the psychological state and academic performance of vocational college students have become the focus of attention for educators. This study aims to construct a correlation model between the positive psychological state and academic performance of vocational college students based on data mining algorithms to offer a conceptual foundation and practical guidance for the optimization of vocational education. The relationship between positive psychological state and academic performance was analyzed through a literature review, as well as the application of data mining algorithms in the field of education. A certain amount of data on vocational college students was collected using questionnaire surveys and empirical research methods, including their basic information, positive psychological status indicators, and academic performance data. Subsequently, data mining algorithms were used to preprocess and analyze the collected data, and a correlation model between the positive psychological state and academic performance of vocational college students was constructed. Finally, through validation and evaluation of the model, it was found that there is a significant positive correlation between positive psychological state and academic performance, and the model has high predictive accuracy. The study's results suggest that the positive psychological state of vocational college students has a significant impact on their academic performance. Educators should consider students' mental health and take effective measures to enhance their positive psychological state, thereby improving their academic performance. This study provides a new research perspective and method for the field of vocational education, which helps to promote the development and reform of vocational education.

Author 1: Zeng Daoyan

Author 2: Chen Disi

Keywords: Data mining algorithms; vocational students; positive psychological state; academic performance; correlation model

Paper 8: Physical Training in Higher Vocational Colleges Based on Sequencing Adaptive Genetic Algorithm

Abstract: This study is based on the sequencing adaptive genetic algorithm and conducts an in-depth discussion on optimization issues in the field of higher vocational sports training. By analyzing the shortcomings of traditional genetic algorithms in optimizing training plans, a new sequencing adaptive genetic algorithm is proposed to improve the optimization effect and adaptability of training plans. First, the optimization goals and constraints in higher vocational sports training were studied, including the diversity of training content and the rationality of training intensity. Secondly, based on the sequencing adaptive genetic algorithm, an optimization algorithm framework suitable for higher vocational sports training was designed, including key steps such as individual coding, fitness evaluation, and crossover mutation. Then, the proposed algorithm was verified and analyzed using experimental data. The results showed that the algorithm can effectively improve the optimization effect of the training plan and has strong adaptability and generalization capabilities. Finally, through comparison with traditional genetic algorithms and other optimization algorithms, the superiority and practicability of sequencing adaptive genetic algorithms in higher vocational sports training are further verified.

Author 1: Quanzhong Gao

Keywords: Sequencing adaptive genetic algorithm; higher vocational colleges; sports training; convergence speed

Paper 9: Packaging Beautification Design Based on Visual Image and Personalized Pattern Matching

Abstract: Visual image technology is widely used in the field of product art design, enriching the visual beautification design effect of products. To improve the design effect of product packaging, a personalized packaging pattern matching technology is proposed based on computer vision image technology. Firstly, based on user needs, a pattern feature extraction technology is proposed, which uses the total variation model and GrabCut model to smooth and segment the image. Secondly, an improved style transfer generative adversarial network model is proposed for transfer training between feature elements and targets. Considering the problem of insufficient detail preservation in traditional transfer models, attention layers are incorporated into the transfer model for improvement. In the pattern feature extraction experiment, the proposed model had the best pixel accuracy in Image 1. In the pattern matching experiment, the proposed model had the lowest mapping loss in both pattern combinations, with a value of 0.135 in the Zhuang brocade pattern and 0.236 in the blue and white porcelain pattern, which was superior to other models. Comparing the effect of different model pattern combinations, in the blue and white porcelain pattern combination, the proposed model had an optimal peak signal-to-noise ratio of 32.32, which was superior to other models. The proposed model has excellent application effects in packaging design beautification. The research content will provide critical technical references for e-commerce product packaging design and intelligent image processing.

Author 1: Deli Chen

Keywords: Visual images; personalized patterns; total variational model; GrabCut model; migration model

Paper 10: Blockchain-based Cannabis Traceability in Supply Chain Management

Abstract: The typical cannabis supply chain is encountering obstacles with the traceability of product regulations and standards. It is a complex structure involving multiple organizations and healthcare products. Questionable products finding their way onto the legal market are potentially dangerous. The proportion of Tetrahydrocannabinol (THC)/Cannabidiol (CBD) and the source of the cannabis strains have an impact on human treatment, limiting the traditional cannabis supply chain from seed to sale. Currently, the cannabis supply chain involves multiple stakeholders, which complicates the validation of various essential criteria, including license management, Certificate of Analysis (COA), and conformance quality standards and regulations. Existing traceability systems involve a centralized authority, leading to a lack of transparency and tracking system immutability. This study offers a Polygon blockchain-based strategy using smart contracts and decentralized on-chain and off-chain storage for efficient information searches in the cannabis supply chain. Eliminating the need for middlemen, the blockchain-based solution gives data security and transaction immutability history to all stakeholders. The storage structure comprises on-chain and off-chain components, algorithms, and the operating principles of the suggested solution. In addition, the suggested system delivers query efficiency and assures supply chain management authenticity and dependability. To assess the performance of the cannabis supply chain, scalability in developing a blockchain-based traceability process avoids delays and high transaction fees.

Author 1: Piwat Nowvaratkoolchai

Author 2: Natcha Thawesaengskulthai

Author 3: Wattana Viriyasitavat

Author 4: Pramoch Rangsunvigit

Keywords: Blockchain; cannabis; traceability; supply chain management; polygon; on-chain and off-chain

Paper 11: Predictive Modeling of Kuwaiti Chronic Kidney Diseases (KCKD): Leveraging Electronic Health Records for Clinical Decision-Making

Abstract: Chronic kidney disease (CDK) represents a significant public health concern globally, and its prevalence is on the rise. In the context of Kuwait, this study addresses the imperative of predicting CKD by leveraging the wealth of information embedded in electronic health records (EHRs). The primary objective is to develop a predictive model capable of early identification of individuals at risk for CKD, thereby enabling timely interventions and personalized healthcare strategies and equip clinicians with information that enhances their ability to make well-informed decisions regarding prognoses or therapeutic interventions. In this study, a dataset has been created from Kuwaiti healthcare institutions, emphasizing the richness and diversity of patient information encapsulated in EHRs and a feature engineering step has been applied for labeling it. Various ensemble learning algorithms, Ada Boost, Extreme Gradient Boosting, Extra Trees, Gradient Boosting, Random Forest, and various single learning algorithms, Decision Tree, K-Nearest Neighbors, Logistic Regression, Multilayer Perceptron, Stochastic Gradient Descent, Support Vector Machines, have been implemented. By examining the empirical findings of our tests, our results showcase the models’ capability to identify individuals at risk for CKD at an early stage, facilitating targeted healthcare interventions. Decision Tree was the best classifier achieving 99.5% accuracy and 99.3% macro averaged f1-score.

Author 1: Talal M. Alenezi

Author 2: Taiseer H. Sulaiman

Author 3: Mohamed Abdelrazek

Author 4: Amr M. AbdelAziz

Keywords: Chronic kidney diseases; Electronic Health Records (EHR); classification; machine learning

Paper 12: Data Manipulation in Wireless Sensor Networks: Enhancing Security Through Blockchain Integration with Proposal Mitigation Strategy

Abstract: In recent years, Wireless Sensor Networks (WSNs) have become integral in various applications ranging from environmental monitoring to defense. However, the security and reliability of these networks remain a paramount concern due to their susceptibility to various types of cyber-attacks and failures. This paper proposes a novel integration of blockchain technology with WSNs to address these challenges. Blockchain, with its decentralized and tamper-resistant ledger, offers a robust framework to enhance the security and reliability of sensor networks. The study begins by analyzing the current security threats and challenges faced by WSNs, emphasizing the need for a solution that can ensure data integrity, confidentiality, and network resilience. We then introduce blockchain technology and discuss its key features such as decentralization, immutability, and consensus algorithms, which are beneficial in creating a secure and reliable WSN environment. Subsequently, we present a detailed architecture of how blockchain can be integrated with WSNs. This includes the deployment of a lightweight blockchain protocol suited for the limited computational resources of sensor nodes. We also explore the use of smart contracts for automated, secure data handling and network management within WSNs. To validate the proposed integration, we conduct a simulations based on network attacks. The results demonstrate significant improvements in the security and reliability of WSNs when blockchain is implemented. This is evidenced by enhanced resistance to common attacks, such as data manipulation and node compromise and increased network uptime.

Author 1: Ayoub Toubi

Author 2: Abdelmajid Hajami

Keywords: Wireless sensor networks; blockchain technology; network security; data integrity

Paper 13: Web-based Expert Bots System in Identifying Complementary Personality Traits and Recommending Optimal Team Composition

Abstract: The use of web-based expert systems in the workplace has become increasingly common in recent years, with companies using these automated tools to streamline a range of tasks, from customer service to employee training. However, the potential of web-based expert bots systems to help build more effective teams by identifying employees with complementary personality traits and providing recommendations for team composition has received less attention. This paper investigates the application of a web-based expert bots’ system in identifying complementary personality traits among employees to recommend optimal team compositions. We developed a web-based expert bot system, augmented by a chatbot interface, to evaluate and synthesize employee personality profiles for improved team alignment. The results, derived from questionnaire feedback and prototype assessments, demonstrate the system's capability to enhance team performance metrics and behavioral competencies. The discussion outlines the system's advantages, and its potential in organizational settings, and acknowledges its limitations. Web-based expert systems with chatbots that exhibit unique personalities tend to be more engaging and effective. Consequently, this system is expected to not only foster better team cohesion but also to increase user involvement and satisfaction. Future work is dedicated to expanding the system's capabilities and conducting extensive field testing to establish its practical effectiveness.

Author 1: Mysaa Fatani

Author 2: Haneen Banjar

Keywords: Web-based expert system; personality traits; team composition; workplace efficiency and chatbot integration

Paper 14: Hybrid Intrusion Detection System Based on Data Resampling and Deep Learning

Abstract: The growth of the internet has advanced information-sharing capabilities and vastly increased the importance of global network security. However, because new and inconspicuous abnormal behaviors are nearly impossible to detect in massive network access environments, modern intrusion detection systems have identified a high rate of false-positive (FP) and false-negative (FN) attacks. To overcome this, this paper proposes a hybrid deep learning model that significantly mitigates the disadvantages of consistently imbalanced sample attack data. First, it resolves imbalanced data using random undersampling and synthetic minority oversampling techniques. Then, convolutional neural networks (CNNs) extract local and spatial features, and a transformer encoder extracts global and temporal features. The novelty of this combination increases recognition accuracy at the algorithm level, which is crucial to reducing FPs and FNs. The model was subjected to multiclassification testing on the NSL-KDD and CICIDS2017 benchmark datasets, and the results show that our model has higher classification accuracy and lower FP rates than state-of-the-art intrusion detection models. Moreover, it significantly improves the detection rate of low-frequency attacks.

Author 1: Huan Chen

Author 2: Gui-Rong You

Author 3: Yeou-Ren Shiue

Keywords: Intrusion detection; deep learning; random undersampling; synthetic minority oversampling technique; convolutional neural network; transformer

Paper 15: Addressing Imbalanced Data in Network Intrusion Detection: A Review and Survey

Abstract: The proliferation of internet-connected devices, including smartphones, smartwatches, and computers, has led to an unprecedented surge in data generation. The rapid rise in device connectivity points to an urgent need for robust cybersecurity measures to counter the mounting wave of cyber threats. Among the strategies aimed at establishing efficient network intrusion detection systems, the integration of machine learning techniques is a prominent avenue. However, the application of machine learning models to imbalanced intrusion detection datasets, such as NSL-KDD, CICIDS2017, and UGR'16, presents challenges. In such intricate scenarios, accurately distinguishing network intrusions poses a formidable challenge. The term "imbalance" refers to the imbalanced distribution of data across classes, which adversely affects the precision of machine learning algorithm classifications. This comprehensive survey embarks on a thorough exploration of the spectrum of methodologies proposed to address the challenge of imbalanced data. Simultaneously, it assesses the efficacy of these methodologies within the realm of network intrusion detection. Moreover, by shedding light on the potential consequences of not effectively tackling imbalanced data, this study aims to provide a holistic understanding of the intricate interplay between machine learning and intrusion detection in imbalanced settings.

Author 1: Elham Abdullah Al-Qarni

Author 2: Ghadah Ahmad Al-Asmari

Keywords: Network intrusion detection system; data imbalance; resampling; data level techniques; hybrid techniques

Paper 16: Roadmap for Generative Models Redefining Learning in Egyptian Higher Education

Abstract: Artificial Intelligence (AI) Generative models have become powerful tools in all sciences, research, academia, and businesses. Egyptian Universities need to leverage those models while using them ethically and responsibly to survive in the current global market. This paper explains the evolution of those models, from basic natural language processing by IBM in 1954 to the current powerful revolutionary generative models. The paper presents research that helps us get desired outputs or behaviors from generative models through prompt engineering, chain of thought prompting and ReAct. The paper presents Egypt and Egyptian Universities readiness and steps taken to get advantage of the latest AI technologies. The paper examines the training of those models to identify their advantages and disadvantages for university members focusing on the Egyptian context. The roadmap for Egyptian Universities use of generative models consists of a SWOT analysis; an infographic of policies and guidelines with regard to faculty and students use of generative models at Egyptian Universities promoting academic integrity and innovation, while minimizing the risks associated with this technology; A table of types and severities of penalties for policy violations by students using generative models is specified and finally a framework for nontechnical users of generative models of reusable patterns to get the optimal desired output of the models is developed.

Author 1: Laila Mohamed ElFangary

Keywords: Artificial intelligence; generative models; prompt engineering; higher education; Egyptian universities

Paper 17: A Smart Framework for Enhancing Automated Teller Machines (ATMs) Fraud Prevention

Abstract: Over the past years, clients have largely depended on and trusted Automated Teller Machines (ATMs) to fulfill their banking needs and control their accounts easily and quickly. Despite the significant advantages of ATMs, fraud has become a very high risk and danger. As it leads to controlling all clients' accounts. In this paper, the proposed framework is using the iris recognition technology combined with the one-time password (OTP) to detect and prevent the known as well as the unknown attacks on ATMs and provide a table of the attackers and the suspected attackers with a counter to take a preventive action with them. Our proposed preventive actions are: card withdrawal, flagging the identified iris as an attacker in the database, notifying the card owner with this suspicious behavior, reporting to the Central Bank of Egypt (CBE), and calling the police when an attacker's iris counts three capturing times, even if for a different card. Two case studies were attempted to achieve the highest accuracy, the first case was using the Chinese Academy of Sciences' Institute of Automation V1.0 (CASIA-IrisV1) dataset using the Cosine Distance. The second one was using the Indian Institute of Technology Delhi (IITD) dataset using k-Nearest Neighbors (KNN) and Histogram of Oriented Gradient (HOG) techniques together reaching 100% accuracy.

Author 1: Mohamed Abdelsalam Ahmed

Author 2: Nada Tarek Abbas Haleem

Author 3: Amira M. Idrees

Keywords: Automated Teller Machines (ATMs); digital banking; image processing; iris recognition; One Time Password (OTP); machine learning; fraud detection; fraud prevention; biometrics; security; banking

Paper 18: A Combined Ensemble Model (CEM) for a Liver Cancer Detection System

Abstract: The liver is one of the most important organs in the human body. The liver's proper function is critical for overall health, and liver diseases or disorders can have serious consequences. Liver cancer is also known as hepatic cancer, which is divided into various types of cells that belong to the cancer. The most common type of liver cancer is hepatocellular carcinoma (HCC). HCC is one of the most common types of liver cancer that can affect up to 85% of people worldwide. Early detection of liver cancer is essential in healthcare because it increases the chances of successful treatment and patient outcomes. Many researchers have developed models that help detect and diagnose liver cancer. The first step in detecting liver cancer is identifying people at a higher risk. Chronic hepatitis B or C infection, cirrhosis, heavy alcohol use, obesity, and exposure to certain chemicals and toxins are all risk factors. This paper is mainly focused on detecting the cancer-affected regions that occur in the liver. In this paper, a combined ensemble model (CEM) for a liver cancer detection system is developed to find and detect liver cancer and liver disorders in their early stages. A pre-trained model, RESNET50 with transfer learning, is used to obtain the features from the pre-trained model—an advanced preprocessing technique involved in filtering the noise from input CT scan images. A hybrid feature extraction (HFE) technique also gets significant elements from the input CT scan images. Finally, the proposed CEM combines an Extreme Gradient Boosting (EGB) algorithm with a Recurrent Neural Network (RNN) that focuses on detecting the abnormal cancer cells present in input CT scan images. The performance of the CEM shows a high accuracy of 98.48% with a 10% high detection rate. Previously, it was 88.12%.

Author 1: T. Sumallika

Author 2: R. Satya Prasad

Keywords: Liver Cancer; Hepatocellular Carcinoma (HCC); Combined Ensemble Model (CEM); RESNET50; Extreme Gradient Boosting (EGB); Recurrent Neural Network (RNN)

Paper 19: Automatic Dust Reduction System: An IoT Intervention for Air quality

Abstract: Air quality is of great importance due to its direct impact on the environment, human health, and quality of life. It could be affected negatively by the presence of dust particles in the atmosphere. Thus, it is vital to purify air from dust and mitigate its impact on air quality. In this regard, dust sensors play a vital role in monitoring and measuring airborne dust particles. They utilize various techniques, such as optical scattering, to detect and quantify the concentration of dust in the air. Microcontrollers are powerful and versatile devices, which have been widely used in many Internet of Things (IoT) applications. They process the collected data from sensors and react accordingly by controlling the operation of IoT devices. Accordingly, the primary goal of this paper is to develop a model for reducing the amount of dust and other particulates in the air to improve its quality. In addition to the microcontroller, which controls the overall operation of the proposed model, two other main components are utilized: a sensor and a sprinkler. The results of the model have shown that it can successfully reduce the dust concentration and suppress the dust intensity to less than 0.1%. The result concluded that the proposed model achieved its primary goals by integrating sensors and sprinkler into an intelligent dust removal model.

Author 1: Bosharah Makki Zakri

Author 2: Ohoud Alzamzami

Author 3: Amal Babour

Keywords: Dust Suppression; dust elimination; digital dust sensor; humidifier; dust intensity

Paper 20: Q-KGSWS: Querying the Hybrid Framework of Knowledge Graph and Semantic Web Services for Service Discovery

Abstract: In the era of big data, Knowledge Graphs (KGs) have become essential tools for managing interconnected datasets across various domains. This paper introduces a novel RDF (Resource Description Framework) based Knowledge Graph of Semantic Web Services (KGSWS), designed to enhance service discovery. Leveraging the versatile SPARQL query language, the framework facilitates precise querying operations on KGSWS, enabling customized service matching for user queries. Through comprehensive experimentation and analysis, notable improvements in accuracy (69.75% and 90.01%) and rapid response times (0.61s and 1.57s) across two semantic search levels are demonstrated, validating the efficacy of the approach. Furthermore, research questions regarding the interlinking of ontologies, methods for formulating automatic queries, and efficient retrieval of services are addressed, offering insights into future avenues for research. This work represents a significant advancement in the domain of semantic web services, with potential applications across various industries reliant on efficient service identification and integration. Future phases of research will focus on logical inference and the integration of machine learning-based graph embedding models, promising even greater strides in knowledge discovery within the KGSWS framework, thus reshaping the domain of semantic web services.

Author 1: Pooja Thapar

Author 2: Lalit Sen Sharma

Keywords: Ontologies; knowledge graph; semantic web services; SPARQL query language; OWLS; data integration; service discovery

Paper 21: Rural Revitalization Evaluation using a Hybrid Method of BP Neural Network and Genetic Algorithm Based on Deep Learning Model

Abstract: The rural revitalization strategy is a comprehensive plan for supporting rural revival in the new development stage while prioritizing agricultural and rural area development. Establishing a rural revitalization evaluation model will help monitor and guide the development of rural revitalization strategies and comprehensively deepen rural reforms. This research combines the benefits of the BP neural network with the genetic algorithm, introduces the genetic algorithm in optimizing the weights and thresholds of the BP neural network, and develops a GA-BP neural network model to evaluate and predict rural rejuvenation. The research findings demonstrated that the GA-BP neural network model possesses rapid convergence, accuracy, and stability in assessing and predicting rural revival and can evaluate and predict rural revitalization well.

Author 1: Songmei Wang

Author 2: Min Han

Keywords: The rural revitalization strategy; deep learning model; the GA-BP neural network; evaluation model

Paper 22: Sound Classification for Javanese Eagle Based on Improved Mel-Frequency Cepstral Coefficients and Deep Convolutional Neural Network

Abstract: The Javanese Eagle is a rare and protected animal in Indonesia. These animals only live in a few species and are threatened with extinction. These birds need to be bred to avoid extinction. One form of communication between the Javanese eagles and each other is the sound of their tweets. These tweets can be studied and classified to conserve endangered animals. This study will classify the sound of the Javanese Eagles for the benefit of animal conservation. Data in the form of voice tweets will be classified. This classification uses algorithms from improved MFCC (Mel-Frequency Cepstral Coefficients) and Deep Convolutional Neural Network. The result of this study was to classify the sound of the Javanese Eagle from the lack of food or drink, the normal tweets state of the bird, and to find out the Javanese Eagle in finding a partner. This research has been carried out by comparing the CNN architecture with AlexNet and VGGNet models and various combinations of training, validation, and test data. The best model dataset underwent division into 80% for training, 10% for validation, and 10% for testing. Training and testing of both IMFCC and VGGNet models occurred using the same dataset. During training, VGGNet achieved 100% accuracy, while testing yielded 99%. ROC Curve: 'Normal' AUC 0.996, 'Looking for Partner' AUC 1.000, 'Looking for Food' AUC 0.996. This study aids Javanese Eagle conservation, crucial for preventing extinction at conservation sites.

Author 1: Silvester Dian Handy Permana

Author 2: T. K. Abdul Rahman

Keywords: Improved MFCC; deep convolutional neural network; Javanese eagle sound; sound classification

Paper 23: Ethnicity Classification Based on Facial Images using Deep Learning Approach

Abstract: Race and ethnicity are terminologies used to describe and categorize humans into groups based on biological and sociological criteria. One of these criteria is the physical appearance such as facial traits which are explicitly represented by a person’s facial structure. The field of computer science has mostly been concerned with the automatic detection of human ethnicity using computer vision-based techniques, where it can be challenging due to the ambiguity and complexity on how an ethnic class can be implicitly inferred from the facial traits in terms of quantitative and conceptual models. The current techniques for ethnicity recognition in the field of computer vision are based on encoded facial feature descriptors or Convolutional Neural Network (CNN) based feature extractors. However, deep learning techniques developed for image-based classification can provide a better end to end solution for ethnicity recognition. This paper is a first attempt to utilize a deep learning-based technique called vision transformer to recognize the ethnicity of a person using real world facial images. The implementation of Multi-Axis Vision Transformer achieves 77.2% classification accuracy for the ethnic groups of Asian, Black, Indian, Latino Hispanic, Middle Eastern, and White.

Author 1: Abdul-aziz Kalkatawi

Author 2: Usman Saeed

Keywords: Vision transformer; deep learning; ethnicity; race; classification; recognition

Paper 24: A Driving Area Detection Algorithm Based on Improved Swin Transformer

Abstract: Drivable area or free space detection is an essential part of the perception system of an autonomous vehicle. It helps intelligent vehicles understand road conditions and determine safe driving areas. Most of the driving area detection algorithms are based on semantic segmentation that classifies each pixel into its category, and recent advances in convolutional neural networks (CNNs) have significantly facilitated semantic segmentation in driving scenarios. Though promising results have been obtained, the existing CNN-based drivable area detection methods usually process one local neighborhood at a time. The locality of convolutional operation fails to capture long-range dependencies. To solve this problem, we propose an improved Swin Transformer based on shift window, named Multi-Swin. First, an improved patch merging strategy is proposed to enhance feature interactions between adjacent patches. Second, a decoder with upsampling layer is designed to restore the resolution of the feature map. Last, a multi-scale fusion module is utilized to improve the representation ability of global semantic and geometric information. Our method is evaluated and tested on the publicly available Cityscapes dataset. The experimental results show that our method achieves 91.92% IoU in road segmentation detection, surpassing state-of-the-art methods.

Author 1: Shuang Liu

Author 2: Ying Li

Author 3: Huankun Sheng

Keywords: CNNS; driving area detection; multiscale fusion; semantic segmentation; Swin Transformer

Paper 25: Sky Pixel Detection in Outdoor Urban Scenes: U-Net with Transfer Learning

Abstract: The sky depicts a high visual importance in outdoor scenes, often appearing in video sequences and photos. Sky information is crucial for accurate sky detection in several computer vision applications, such as scene understanding, navigation, surveillance, and weather forecasting. The difficulty of detecting is clarified by variations in the sky's size, weather and lighting conditions, and the sky's reflection on other objects. This article presents a new contribution to address the challenges facing sky detection. A unique dataset was built that includes scenes of distinct lighting and atmospheric phenomena. Additionally, a modified U-Net architecture was proposed with pre-trained models as encoder VGG19, EfficientNetB4, InceptionV3, and DenseNet121 for sky detection to solve outdoor image limitations and evaluate the influence of different encoders when integrated with the U-Net, aiming to identify which encoder describes features of the sky accurately. The proposed approach shows encouraging results; as it presents improved performance over the adjusted U-Net architecture with inceptionv3 on the proposed dataset, achieving mean Intersection over union, dice similarity coefficient, recall, precision, and accuracy of 98.57 %, 99.57 %, 99.41 %, 99.73%, and 99.40 %, respectively. At the same time, the best loss was achieved in U-Net with VGG19 equivalent of 0.09.

Author 1: Athar Ibrahim Alboqomi

Author 2: Rehan Ullah Khan

Keywords: Computer vision; transfer learning; semantic segmentation; sky detection; U-Net; machine learning

Paper 26: Revolutionizing Plant Disease Detection in Leaves: An Innovative Hybrid ABOCNN Framework for Advanced and Accurate Identification

Abstract: Plant diseases are a persistent threat to the global agricultural economy, compromising food supply and security. Accurate and early diagnosis is vital for effective agricultural management. This study addresses this gap by introducing a better approach for identifying plant diseases in leaves: the Integrated Hybrid Attention-Based One-Class Neural Network (ABOCNN) System. The system uses deep learning and domain-specific information, as well as powerful neural networks and attention processes, to extract features unique to a certain ailment while excluding irrelevant data. By dynamically focusing on prominent areas in leaf images, the proposed methodology obtains an impressive 99.6% accuracy, beating both traditional approaches and cutting-edge deep-learning approaches by an average of 12.7%. The practical use of this strategy has a significant influence on crop yield and agricultural sustainability. Attention maps increase interpretability and help individuals comprehend more fully how decisions are made. The system, written in Python, is precise, scalable, and adaptable, making it a helpful tool for a wide range of agricultural applications combining multiple plant species and disease classifications. With an incredible 99.6% accuracy rate, the Integrated Hybrid ABOCNN Technology provides an innovative method for diagnosing plant diseases, outperforming conventional approaches by 12.7%. Attention maps increase interpretability and give important information about the model's decision-making processes.

Author 1: V. Krishna Pratap

Author 2: N. Suresh Kumar

Keywords: Convolutional Neural Network (CNN); attention model; leaf disease detection; attention-based one-class neural network; crop production

Paper 27: AI-Enhanced Comprehensive Liver Tumor Prediction using Convolutional Autoencoder and Genomic Signatures

Abstract: Liver tumor prediction plays a pivotal role in optimizing treatment strategies and improving patient outcomes. In our proposed work, we present an innovative AI-driven framework for liver tumor prediction, uniting cutting-edge techniques to enhance precision and depth of analysis. The framework integrates a Histological Convolutional Autoencoder (HistoCovAE) for meticulous tumor segmentation in medical imaging, and Genomic Feature Extraction (MIRSLiC) for a nuanced understanding of molecular markers. Additionally, a Multidimensional Feature Extraction module amalgamates videomics, radiomics, acoustics, and clinical data, creating a comprehensive dataset. These dimensions synergize in a unified model, offering detailed predictions encompassing tumor characteristics, subtypes, and prognosis. Model evaluation and continuous improvement, guided by real-world outcomes, underscore reliability. This integrative approach transcends conventional boundaries, providing clinicians’ actionable insights for personalized treatment strategies and heralding a new era in liver tumor prediction. Our model undergoes rigorous evaluation against diverse datasets, and the performance metrics underscore its reliability and accuracy. With precision exceeding 87%, recall rates above 92%, and a Dice coefficient surpassing 0.89 in tumor segmentation, our model showcases exceptional accuracy and robustness. In prognostic modeling, survival prediction accuracy consistently surpasses 84%, highlighting the model's ability to provide valuable insights into the future trajectory of liver cancer.

Author 1: G. Prabaharan

Author 2: D. Dhinakaran

Author 3: P. Raghavan

Author 4: S. Gopalakrishnan

Author 5: G. Elumalai

Keywords: Liver tumor prediction; autoencoder; segmentation; feature extraction; genomics; artificial intelligence

Paper 28: Improved ORB Algorithm Through Feature Point Optimization and Gaussian Pyramid

Abstract: Feature points obtained using traditional ORB methods often exhibit redundancy, uneven distribution, and lack scale invariance. This study enhances the traditional ORB algorithm by presenting an optimal technique for extracting feature points, thereby overcoming these challenges. Initially, the image is partitioned into several areas. The determination of the quantity of feature points to be extracted from each region takes into account both the overall number of feature points and the number of divisions that the image undergoes. This method tackles concerns related to the overlap and redundancy of feature points in the extraction process. To counteract the non-scale invariance issue in feature points obtained via the ORB method, a Gaussian pyramid is employed, and feature points are extracted at each level. Experimental findings demonstrate that our method successfully extracts feature points with greater uniformity and rationality, while preserving image matching accuracy. Specifically, our technique outperforms the traditional ORB algorithm by approximately 4% and the SURF algorithm by 2% in terms of matching performance. Additionally, the processing time of our proposed algorithm is three times faster than that of the SURF algorithm and twelve times faster than the SIFT algorithm.

Author 1: Rohmat Indra Borman

Author 2: Agus Harjoko

Author 3: Wahyono

Keywords: Feature point; Gaussian pyramid; image matching; ORB algorithm; scale invariance

Paper 29: Performance-Optimised Design of the RISC-V Five-Stage Pipelined Processor NRP

Abstract: The five-stage pipeline processor is a mature and stable processor architecture suitable for many applications in the field of computer hardware. Based on the RISC-V instruction set architecture, the five-stage pipeline processor has advantages in performance, functionality, and power consumption. This paper presents an optimized RV32I five-stage pipeline processor, NRP, and proposes two optimization methods to improve the performance of NRP. These methods include instruction decoding unit optimization and branch prediction optimization. We implemented NRP using Verilog HDL and verified its performance using Vivado and the Xilinx Artya7-35T FPGA board. Experimental data shows that after adopting these methods, the CoreMark score of the five-stage pipeline processor reached 3.11 CoreMark/MHz, representing an 11.07% performance improvement.

Author 1: Hongkui Li

Author 2: Chaoxia Jing

Author 3: Jie Liu

Keywords: Architecture; FPGA; RISC-V; RV32I; Verilog HDL; five-stage

Paper 30: Enhancing Thyroid Cancer Diagnostics Through Hybrid Machine Learning and Metabolomics Approaches

Abstract: Thyroid cancer, a prevalent endocrine malignancy, necessitates advanced diagnostic techniques for accurate and early detection. This study introduces an innovative approach that integrates hybrid Machine Learning (ML) algorithms with metabolomics, offering a novel pathway in thyroid cancer diagnostics. Our methodology employs a range of hybrid ML models, combining the strengths of various algorithms to analyze complex metabolomic data effectively. These models include ensemble methods, neural network-based hybrids, and integrations of unsupervised and supervised learning techniques, tailored to decipher the intricate patterns within metabolic profiles associated with thyroid cancer. The study demonstrates how these hybrid ML algorithms can efficiently process and interpret metabolomic data, leading to enhanced diagnostic accuracy. By leveraging the distinct characteristics of each ML model, our approach not only improves the detection of thyroid cancer but also contributes to a deeper understanding of its metabolic underpinnings. The findings of this study pave the way for more personalized and precise medical interventions in thyroid cancer management, showcasing the potential of hybrid ML models in revolutionizing cancer diagnostics. Our system analyzes thyroid cancer metabolomic data using ensemble methods, neural network-based hybrids, and unsupervised and supervised learning integrations. The research shows hybrid ML models may revolutionize cancer diagnoses by improving accuracy. LSTM+CNN, LSTM+GRU, and CNN+GRU have high accuracy rates, helping us comprehend thyroid cancer's biochemical roots. Hybrid ML models enhance thyroid cancer diagnosis and management, enabling more tailored and accurate medical treatments. The hybrid machine learning models like LSTM+CNN, LSTM+GRU, and CNN+GRU beat CNN, VGG-19, Inception-ResNet-v2, decision support, and random forests (99.45%).

Author 1: Meghana G Raj

Keywords: Thyroid cancer; hybrid ML models; metabolomics; diagnostic accuracy

Paper 31: Precision Insulin Delivery: Predictive Modelling for Bolus Insulin Injection in Real-Time

Abstract: Insulin is recommended for patients with Diabetes Mellitus (DM). It is challenging for doctors to prescribe accurate bolus insulin before every meal due to real-time factors such as the size of the meal, skipping a previous meal, and physical activity, which can risk the patient towards hyperglycemia or hypoglycemia. Previous studies executed insulin predictions where the methods did not consider the cases of controlled glucose levels, type of insulin prescribed, time of insulin-induced, and data detersion that can alter the predictions. To address these problems, our work has proposed an insulin predictive model from the integration of Internet of Things (IoT) devices, i.e., Continuous glucose monitoring (CGM) sensor and insulin pumps with rapid-acting insulin type where the insulin dosage with corresponding Current Blood Glucose levels (CBG) and improved Next Blood Glucose levels (NBG) are chosen. The dataset is subjected to data detersion where pre-processing, Exploratory Data Analysis (EDA), and Feature Selection is performed. Machine Learning (ML) models are applied on curated dataset where Decision Tree (DT)-Bagging algorithm, performed the best with a Mean Absolute Error (MAE) of 1.54 and a Mean Square Error (MSE) of 4.15. Performance metrics of the current study imply its suitability in medical applications for accurate prediction of real-time insulin dosage.

Author 1: V. K. R. Rajeswari Satuluri

Author 2: Vijayakumar Ponnusamy

Keywords: Continuous glucose monitoring; bolus insulin prediction; data curation; data detersion; diabetes mellitus; exploratory data analysis; feature selection; machine learning; pre-processing

Paper 32: Cross-Modal Video Retrieval Model Based on Video-Text Dual Alignment

Abstract: Cross-modal video retrieval remains a major challenge in natural language processing due to the natural semantic divide between video and text. Most approaches use a single encoder to extract video and text features separately, and train video-text pairs by means of contrastive learning, but this global alignment of video and text is prone to neglecting more fine-grained features of both. In addition, some studies focus only on profiling the video description text, ignoring the correlation relationship with the video. Therefore, this paper proposes a video retrieval method based on video-text alignment, which realizes both global and fine-grained alignment between video and text. For global alignment, the video and text are aligned by a single encoder and after linear projection; for fine-grained alignment, the video encoder is trained to align the video and text by masking some semantic information in the text. By experimentally comparing with multiple existing methods on MSR-VTT and MSVD datasets, the model achieves R@1 (recall at 1) metrics of 51.5% and 52.4% on MSR-VTT and MSVD datasets, respectively, which indicates that the proposed model can improve the efficiency of cross-modal video retrieval.

Author 1: Zhanbin Che

Author 2: Huaili Guo

Keywords: Video-text alignment; cross-modal; contrastive learning; similarity measure; feature fusion

Paper 33: Elevating Neuro-Linguistic Decoding: Deepening Neural-Device Interaction with RNN-GRU for Non-Invasive Language Decoding

Abstract: Exploring innovative pathways for non-invasive neural communication with language interfaces, this research delves into the interdisciplinary realm of neurolinguistic learning, merging neuroscience and machine learning. It scrutinizes the intricacies of decoding neural patterns associated with language comprehension. Leveraging advanced neural network architectures, specifically Deep Recurrent Neural Networks (RNN) and Gated Recurrent Units (GRU), the study aims to amplify the landscape of neuro-device interaction. The focus of Neurolinguistic Learning lies in extracting language-related brain signals without resorting to invasive procedures. Employing cutting-edge non-invasive methods and deep learning techniques, the research aims to elevate the capabilities of neural devices such as brain-machine interfaces and neuroprosthetics. A distinctive approach involves crafting a sophisticated Deep RNN-GRU model designed to capture intricate brain patterns linked to language processing. This architectural innovation, implemented in the Python software environment, harnesses the strengths of RNNs and GRUs to enhance language decoding. The study's outcomes hold promise for advancing non-invasive brain language decoding systems, contributing to the expanding knowledge base in neurolinguistic learning. The remarkable accuracy of the proposed RNN-GRU model, boasting a 90% accuracy rate, signifies its potential application in critical real-world scenarios. This includes assistive technologies and brain-machine interfaces where precise decoding of cerebral language signals is paramount. The research underscores the efficacy of deep learning methodologies in pushing the boundaries of neurotechnology. Notably, the model outperforms established techniques, surpassing alternatives like CSP-SVM and EEGNet by an impressive 30.4% in accuracy. The model's proficiency in deciphering topic words underscores its ability to extract intricate language patterns from non-invasive brain inputs.

Author 1: V Moses Jayakumar

Author 2: R. Rajakumari

Author 3: Kuppala Padmini

Author 4: Sanjiv Rao Godla

Author 5: Yousef A.Baker El-Ebiary

Author 6: Vijayalakshmi Ponnuswamy

Keywords: Recurrent Neural Networks (RNN); Gated Recurrent Units (GRU); neurolinguistic learning; neural devices; brain machine interfaces

Paper 34: Post Pandemic Tourism: Sentiment Analysis using Support Vector Machine Based on TikTok Data

Abstract: The tourism industry is one of the hard hit businesses during the Covid-19 pandemic and has been struggling for backup ever since. However, nowadays the industry has started to bloom again with the lifting of all of the restrictions of Covid-19. This research aims to analyze the sentiments of the tourists using the Support Vector Machine (SVM) algorithm to know their views on the tourist spots after the pandemic. The scope of the research covers the state of Terengganu which is popularly known for its islands and unique culture on the east coast of Malaysia. TikTok data has been used as the source of data as social media currently has become one of the top mediums for reviewing, selling and promoting products and services. The objective of the research is to explore the SVM algorithm in the sentiment classification of tourist spots in Terengganu. This research is expected to help the Tourism Terengganu to improve their tourist spots and their services. The phases of the research include collecting data from TikTok, data pre-processing, data labelling, feature extraction, model creation using SVM, graphical user interface development and performance evaluation. The evaluation results showed that the performance of the SVM classifier model was good and reliable, with 90.68% accuracy. The future work would be collecting more data from TikTok regularly to further improve the accuracy of the algorithm.

Author 1: Norlina Mohd Sabri

Author 2: Siti Nur Athira Muhamad Subki

Author 3: Ummu Fatihah Mohd Bahrin

Author 4: Mazidah Puteh

Keywords: Post pandemic tourism; support vector machine; sentiment classification; TikTok data

Paper 35: Investigating the Impact of Train / Test Split Ratio on the Performance of Pre-Trained Models with Custom Datasets

Abstract: The proper allocation of data between training and testing is a critical factor influencing the performance of deep learning models, especially those built upon pre-trained architectures. Having the suitable training set size is an important factor for the classification model’s generalization performance. The main goal of this study is to find the appropriate training set size for three pre-trained networks using different custom datasets. For this aim, the study presented in this paper explores the effect of varying the train / test split ratio on the performance of three popular pre-trained models, namely MobileNetV2, ResNet50v2 and VGG19, with a focus on image classification task. In this work, three balanced datasets never seen by the models have been used, each containing 1000 images divided into two classes. The train / test split ratios used for this study are: 60-40, 70-30, 80-20 and 90-10. The focus was on the critical metrics of sensitivity, specificity and overall accuracy to evaluate the performance of the classifiers under the different ratios. Experimental results show that, the performance of the classifiers is affected by varying the training / testing split ratio for the three custom datasets. Moreover, with the three pre-trained models, using more than 70% of the dataset images for the training task gives better performance.

Author 1: Houda Bichri

Author 2: Adil Chergui

Author 3: Mustapha Hain

Keywords: Artificial intelligence; classification; MobileNetV2; ResNet50v2; sensitivity; specificity; train / test split ratio; VGG19

Paper 36: Action Recognition Method of Basketball Training Based on Big Data Technology

Abstract: Aiming at the problem that improper posture of basketball players leads to not obvious sports effects, the present paper proposes an action recognition method combining computer vision and big data technology and applies it to athletes' daily training and competition. Firstly, based on the current mainstream motion recognition models, 3D graph convolution are used to improve the original 3D convolution to promote the expression ability of spatial structure features and temporal features in skeleton sequences. Secondly, channel and spatial attention mechanisms are introduced to focus on the weight distribution of key points and strong features in different posture recognition processes. Finally, the proposed model is tested in real data, and the test results show that the model runs smoothly while maintaining high recognition performance. It can more effectively direct basketball players to implement comprehensive, systematic, and scientific teaching and training standards that directly support raising the game's general level of performance.

Author 1: Dongsheng CHEN

Author 2: Zhen Ni

Keywords: Action recognition; computer vision; big data technology; three-dimensional convolution; channel and spatial attention mechanisms

Paper 37: Explainable Multistage Ensemble 1D Convolutional Neural Network for Trust Worthy Credit Decision

Abstract: Banking is a dynamic industry that places significant importance on risk management, requiring accurate and interpretable AI models to make transparent lending decisions. This study introduces a groundbreaking approach that combines a multistage ensemble technique with a 1D convolutional neural network (CNN) architecture. The algorithm not only delivers superior classification performance but also offers interpretable explanations for its decisions. The algorithm is designed with multiple strategic steps to enhance model performance without sacrificing explainability. Thorough experiments were conducted using datasets from private banks and non-banking financial companies (NBFCs) in India to evaluate the algorithm's performance. It was compared against various state-of-the-art models, demonstrating remarkable precision, recall, F1 score, and accuracy values of 0.994, 0.992, 0.993, and 0.991, respectively. This outperformed competing models like homogeneous deep ensembles, 1D CNN, and Artificial Neural Networks (ANN). Furthermore, individual borrower dataset evaluations confirmed the proposed algorithm's consistency and efficiency, achieving precision, recall, F1 score, and accuracy values of 0.960, 0.961, 0.952, and 0.964, respectively. The research emphasizes the effectiveness of the explanatory integration decision process, wherein the Explainable Multistage Ensemble 1D CNN not only provides enhanced credit risk prediction but also facilitates transparent and interpretable lending decisions. The algorithm's ability to offer understandable explanations empowers financial institutions to make well-informed lending decisions, reduce credit risk, and foster a more stable and inclusive financial ecosystem.

Author 1: Pavitha N

Author 2: Shounak Sugave

Keywords: Credit risk prediction; explainable AI; multistage ensemble; 1D convolutional neural network; interpretability; transparency; lending decisions; financial institutions

Paper 38: A New Weighted Ensemble Model to Improve the Performance of Software Project Failure Prediction

Abstract: The development of a software project is frequently influenced by various risk factors that can lead to project failure. Predicting potential software project failures early can aid organizations in making decisions regarding possible solutions and improvements. This paper proposes a software project failure prediction model based on a weighted ensemble learning approach. The proposed model aims to determine the failure probability as well as the expected project outcome (Success/Failure). Various ensemble approaches, such as simple majority voting, can be employed in predicting software project failure. However, in majority voting algorithms, all base models have the same weights, resulting in an equal effect on the final prediction result, regardless of their predictive abilities. Our proposed algorithm assigns higher weights to base models that demonstrate a greater ability to correctly predict more challenging data instances. The proposed model is developed based on a dataset gathered from real previous software project reports, comprising both successful and failed projects, to provide evidence supporting the predictive model's capabilities and to obtain high-confidence results. The performance of the developed model is comprehensively assessed through various measures, revealing its superiority in predicting software project failures compared to both simple majority voting and individual models. This research suggests that the proposed model can be integrated into the software system development process, spanning requirement analysis, planning, design, and implementation phases, to evaluate the project's status and identify potential risks.

Author 1: Mohammad A. Ibraigheeth

Author 2: Aws I. Abu Eid

Author 3: Yazan A. Alsariera

Author 4: Waleed F. Awwad

Author 5: Majid Nawaz

Keywords: Ensemble learning; failure prediction; base models; project outcome

Paper 39: Detection of Personal Protective Equipment (PPE) using an Anchor Free-Convolutional Neural Network

Abstract: In industrial environments, the utilization of Personal Protective Equipment (PPE) is paramount for safeguarding workers from potential hazards. While various PPE detection methods have been explored in the literature, deep learning approaches have consistently demonstrated superior accuracy in comparison to other methodologies. However, addressing the pressing research challenge in deep learning-based PPE detection, which pertains to achieving high accuracy rates, non-destructive monitoring, and real-time capabilities, remains a critical need. To address this challenge, this study proposes a deep learning model based on the Yolov8 architecture. This model is specifically designed to meet the rigorous demands of PPE detection, ensuring accurate results. The methodology involves the creation of a custom dataset and encompasses rigorous training, validation, and testing processes. Experimental results and performance evaluations validate the proposed method, illustrating its ability to achieve highly accurate results consistently. This research contributes to the field by offering an effective and robust solution for PPE detection in industrial environments, emphasizing the paramount importance of accuracy, non-destructiveness, and real-time capabilities in ensuring workplace safety.

Author 1: Honggang WANG

Keywords: PPE detection; deep learning; YOLOv8; industrial environments; real-time detection

Paper 40: DeepBiG: A Hybrid Supervised CNN and Bidirectional GRU Model for Predicting the DNA Sequence

Abstract: Understanding the deoxyribonucleic acid (DNA) sequence is a major component of bioinformatics research. The amount of biological data increases tremendously. Hence, there is a need for effective approaches to handle the critical problem in the general computational framework of DNA sequence pre-diction and classification. Numerous deep learning languages can be used to complete these tasks compared to manual techniques that have been followed for ages. The aim of this project is to employ effective approaches for pre-processing DNA sequences and using deep learning languages to train the sequences for making judgments, predictions, and classifications of DNA se-quences into known categories. In this study, the pre-processing methods include k-mers and tokenization. We employ a novel hybrid deep learning algorithm that combines convolutional neural networks and is followed by bidirectional gated recurrent networks. This combination can capture dependencies within the genome sequence, even in large datasets with a lot of noise. The proposed model is compared with existing widely used models and classifiers. The results show that the proposed model achieves a good result with an accuracy of 82.90%. The dataset consists of 44,391 labeled DNA sequences obtained from the Encode project.

Author 1: Chai Wen Chuah

Author 2: Wanxian He

Author 3: De-Shuang Huang

Keywords: DNA sequencing; deep learning; convolutional neural networks; bidirectional gated recurrent; k-mer; tokenizing

Paper 41: Actor Critic-based Multi Objective Reinforcement Learning for Multi Access Edge Computing

Abstract: In recent times, large applications that need near real-time processing are increasingly being used on devices with limited resources. Multi access edge computing is a computing paradigm that provides a solution to this problem by placing servers as close to resource constrained devices as possible. However, the edge device must consider multiple conflicting objectives, viz., energy consumption, latency, task drop rate and quality of experience. Many previous approaches optimize on only one objective or a fixed linear combination of multiple objectives. These approaches don’t ensure best performance for applications that run on edge servers, as there is no guarantee that the solution obtained by these approaches lies on the pareto-front. In this work, Multi Objective Reinforcement Learning with Actor-Critic model is proposed to optimize the drop rate, latency and energy consumption parameters during offloading decision. The model is compared with MORL-Tabular, MORL-Deep Q Network and MORL-Double Deep Q Network models. The proposed model outperforms all the other models in terms of drop rate and latency.

Author 1: Vishal Khot

Author 2: Vallisha M

Author 3: Sharan S Pai

Author 4: Chandra Shekar R K

Author 5: Kayarvizhy N

Keywords: Edge computing; reinforcement learning; multi objective optimization; neural networks; deep learning

Paper 42: A Review on Applications of Electroencephalogram: Includes Imagined Speech

Abstract: In the last two decades, the Brain-Computer Interface system with EEG signals has assisted people in various ways. In particular, to patients with paralysis, epilepsy, and Alzheimer's disease, not only to the patient but also to physically, visually challenged people and Hard-of-Hearing people. One of the non-invasive methods that can read human brain activities is Electroencephalogram (EEG). The EEG has been used in many applications, especially in medicine. The applications of the EEG are not limited to the medical domain; it keeps extending to many areas. This review includes the various application of EEG; and more in imagined speech. The main objective of this survey is to know about imagined speech, and perhaps to some extent, will be useful future direction in decoding imagined speech. Imagined speech classifications have used different models; the models are discussed, the significance of choosing the number of electrodes, and the main challenges in EEG.

Author 1: S. Santhakumari

Author 2: Kamalakannan. J

Keywords: Electroencephalogram; brain signals; invasive; non-invasive; imagined speech; electrodes; epilepsy

Paper 43: Edge Detail Preservation Technique for Enhancing Speckle Reduction Filtering Performance in Medical Ultrasound Imaging

Abstract: Ultrasound imaging is a unique medical imaging modality due to its clinical versatility, manageable biological effects, and low cost. However, a significant limitation of ultrasound imaging is the noisy appearance of its images due to speckle noise, which reduces image quality and hence makes diagnosis more challenging. Consequently, this problem received interest from many research groups and many methods have been proposed for speckle suppression using various filtering techniques. The common problem with such methods is that they tend to distort the edge detail content within the image and blurring is commonly encountered. In this work, we propose a new method that could be combined with previous speckle suppression techniques to preserve edge detail content of the image. The original image is first processed to extract the edge detail content. Rather than presenting the original method to the speckle suppression filtering technique, the edge detail content is subtracted from the original image before it is filtered. Then, such edge detail content is added to the output of filtering to form the final image. The new method is practically verified using 26 imaging experiments as well as ultrasound images from publicly available databases in combination with four widely used speckle reduction filters. The results are evaluated qualitatively and quantitatively using standard image quality metrics.

Author 1: Yasser M. Kadah

Author 2: Ahmed F. Elnokrashy

Keywords: Edge detail preservation; image quality metrics; speckle reduction; ultrasound imaging

Paper 44: A Neuro-Genetic Security Framework for Misbehavior Detection in VANETs

Abstract: Genetic Algorithm (GA) is an excellent optimization algorithm which has attracted the attention of researchers in various fields. Many papers have been published on works done on GA, but no single paper ever utilized this algorithm for misbehavior detection in VANETs. This is because GA requires manual definition of fitness function and defining a fitness function for VANETs is a complex task. Automating the creation of these fitness functions is still a difficulty, even though studies have found several successful applications of GA. In this study, a neuro-genetic security framework has been built with ANN classifier for detecting misbehavior in VANETs. It leverages a genetic algorithm for feature reduction with ANN as a dynamic fitness function, considering both node behaviors and contextual GPS data. Deployed at the Roadside Unit (RSU) level, the framework detects misbehaving nodes, broadcasting alerts to RSUs, Central Authority and the vehicles. The ANN based fitness function has been employed in GA that enabled the GA to select the best results. The 10- fold CV used enabled the whole system to be unbiased giving a precision accuracy of 0.9976 with recall and F1 scores as 0.9977, and 0.9977 respectively. Comparative evaluations, using the VeReMi Extension dataset, demonstrate the framework's superiority in precision, recall, and F1 score for binary and multiclass classification. This hybrid genetic algorithm with ANN fitness function presents a robust, adaptive solution for VANET misbehavior detection. Its context-aware nature accommodates dynamic scenarios, offering an effective security framework for the evolving threats in vehicular environments.

Author 1: Ila Naqvi

Author 2: Alka Chaudhary

Author 3: Anil Kumar

Keywords: VANET security; genetic algorithm; ANN fitness function; misbehavior detection; hybrid detection

Paper 45: Method for Predictive Trend Analytics with SNS Information for Marketing

Abstract: A method for predictive trend analytics with social media information is proposed for marketing. Through keyword analysis, page view analysis, access analysis, heat map analysis, Google Analytics, real time analysis, company and competitor analysis, trend analysis with the social media data derived from X (former tweeter), Instagram, Facebook, YouTube, TikTok, market trend can be predicted. The proposed method is created in a local server and is extended to AWS cloud. The proposed system, also ensure negative / positive analysis from the acquired social media information. Through some experiments, it is found that by using AI to analyze social data by category, you can visualize the degree of attention for each keyword, model relationships between information, identify trending keywords, and where the keywords are in their lifecycle. It turns out that it's possible to categorize which ones exist and predict which ones will scale up in the next six months. In addition, corporate product development and marketing personnel can identify themes, materials, benefits, etc. that have signs of becoming popular based on insights based on predictive behavioral data obtained from the proposed method and system and utilize them in new business development and new product planning.

Author 1: Kohei Arai

Author 2: Ikuya Fujikawa

Author 3: Yusuke Nakagawa

Author 4: Sayuri Ogawa

Keywords: X (former tweeter); Instagram; Facebook; YouTube; TikTok; market trend; AWS; Google analytics; keyword analysis; page view analysis; access analysis; heat map analysis

Paper 46: Study on the Implementation of Multimodal Continuous Authentication in Smartphones: A Systematic Review

Abstract: Profound societal shifts result from the inception of the 4.0 age of the Industrial Revolution and rapid technological advancements. The widespread adoption of e-services has resulted in substantial reliance on smartphones to access diverse offerings. Even so, account breaches and data leaks are risks that users take when they rely so heavily on their smartphones. Authentication is an essential method of safeguarding personal information. The purpose of this study is to undertake a thorough review of the literature on the deployment and trends of multimodal biometric authentication on smartphones. The studies will look at several biometric modalities, such as behavioral and physiological characteristics, and the algorithms for pattern recognition used in continuous authentication systems. The results show various biometric authenticators and emphasize the importance of behavioral features in smartphone authentication. In addition, the research underlines the significance of machine learning algorithms in pattern identification for rapid and accurate analysis. This study helps to understand the present authentication technique landscape and gives ideas for future advances in safe and user-friendly smartphone authentication systems.

Author 1: Rahmad Syalevi

Author 2: Aji Prasetyo

Author 3: Rizal Fathoni Aji

Keywords: Authentication; continuous multimodal; biometric authenticator; smartphone

Paper 47: Elevating Student Performance Prediction using Extra-Trees Classifier and Meta-Heuristic Optimization Algorithms

Abstract: In the highly competitive landscape of academia, the study addresses the multifaceted challenge of analyzing voluminous and diverse educational datasets through the application of machine learning, specifically emphasizing dimensionality reduction techniques. This sophisticated approach facilitates educators in making data-informed decisions, providing timely guidance for targeted academic improvement, and enhancing the overall educational experience by stratifying individuals based on their innate aptitudes and mitigating failure rates. To fortify predictive capabilities, the study employs the robust Extra-Trees Classifier (ETC) model for classification tasks. This model is enhanced by integrating the Gorilla Troops Optimizer (GTO) and Reptile Search Algorithm (RSA), cutting-edge optimization algorithms designed to refine decision-making processes and improve predictive precision. This strategic amalgamation underscores the research's commitment to leveraging advanced machine learning and bio-inspired algorithms to achieve more accurate and resilient student performance predictions in the mathematics course, ultimately aiming to elevate educational outcomes. Analyses of G1 and G3 showcase the efficacy of the ETRS model, demonstrating 97.5% Accuracy, F1-Score, and Recall in predicting the G1 values. Similarly, the ETRS model emerges as the premier predictor for G3, attaining 95.3% Accuracy, Recall, and F1-Score, respectively. These outcomes underscore the significant contributions of the proposed models in advancing precision and discernment in student performance prediction, aligning with the overarching goal of refining educational outcomes.

Author 1: Yangbo Li

Author 2: Mengfan He

Keywords: Student performance; mathematics; machine learning; Extra-Trees Classifier; Gorilla Troops Optimizer; Reptile Search Algorithm

Paper 48: Real-Time Airborne Target Tracking using DeepSort Algorithm and Yolov7 Model

Abstract: In light of the explosive growth of drones, it is more critical than ever to strengthen and secure aerial security and privacy. Drones are used maliciously by exploiting some gaps in artificial intelligence and cybersecurity. Airborne target detection and tracking tasks have gained paramount importance in various domains, encompassing surveillance, security, and traffic management. As airspace security systems aiming to regulate drone activities, anti-drones leverage mostly artificial intelligence and computer vision advances in the used detection and tracking models to perform effectively and accurately airborne target detection, identification, and tracking. The reliability of the anti-drone systems relies mostly on the ability of the incorporated models to satisfy an optimal compromise between speed and performance in terms of inference speed and used detection evaluation metrics since the system should recognize the targets effectively and rapidly to take appropriate actions regarding the target. This research article explores the efficacy of DeepSort algorithm coupled with YOLOv7 model in detecting and tracking five distinct airborne targets namely, drones, birds, airplanes, daytime frames, and buildings across diverse contexts. The used DeepSort and Yolov7 models aim to be used in anti-drone systems to detect and track the most encountered airborne targets to reinforce airspace safety and security. The study conducts a comparative analysis of tracking performance under different scenarios to evaluate the algorithm's versatility, robustness, and accuracy. The experimental results show the effectiveness of the proposed approach.

Author 1: Yasmine Ghazlane

Author 2: Ahmed El Hilali Alaoui

Author 3: Hicham Medomi

Author 4: Hajar Bnouachir

Keywords: Real-time detection; target tracking; anti-drone; Artificial Intelligence; Computer Vision

Paper 49: Towards High Quality PCB Defect Detection Leveraging State-of-the-Art Hybrid Models

Abstract: The automatic detection of defects in printed circuit boards (PCBs) is a critical step in ensuring the reliability of electronic devices. This paper introduces a novel approach for PCB defect detection. It incorporates a state-of-the-art hybrid architecture that leverages both convolutional neural networks (CNNs) and transformer-based models. Our model comprises three main components: a Backbone for feature extraction, a Neck for feature map refinement, and a Head for defect prediction. The Backbone utilizes ResNet and Bottleneck Transformer blocks, which are proficient at highlighting small defect features and overcoming the shortcomings of previous models. The Neck module, designed with Ghost Convolution, refines feature maps. It reduces computational demands while preserving the quality of feature representation. This module also facilitates the integration of multi-scale features, essential for accurately detecting a wide range of defect sizes. The Head employs a Fully Convolutional One-stage detection approach, allowing for the prediction process to proceed without reliance on predefined anchors. Within the Head, we incorporate the Wise-IoU loss to refine bounding box regression. This optimizes the model's focus on high-overlap regions and mitigates the influence of outlier samples. Comprehensive experiments on standard PCB datasets validate the effectiveness of our proposed method. The results show significant improvements over existing techniques, particularly in the detection of small and subtle defects.

Author 1: Tuan Anh Nguyen

Author 2: Hoanh Nguyen

Keywords: PCB defect detection; hybrid neural network; bottleneck transformer; ghost convolution; wise-IoU loss

Paper 50: Advancing Parkinson's Disease Severity Prediction using Multimodal Convolutional Recursive Deep Belief Networks

Abstract: Parkinson's disease (PD), a progressive neurological ailment predominantly affecting individuals over the age of 60, involves the gradual loss of dopamine-producing neurons. The challenges associated with the subjectivity, resource intensity, and limited efficacy of current diagnostic methods, including the Unified Parkinson’s Disease Rating Scale (UPDRS), neuroimaging, and genetic analysis, underscore the need for innovative approaches. This paper introduces a groundbreaking multimodal deep learning framework that integrates Recurrent Neural Networks (RNN-DBN) for precise feature selection and Convolutional Neural Networks (CNNs) for robust feature extraction, aiming to enhance the accuracy of PD severity prediction. The methodology synergistically incorporates genetic data, imaging data from MRI and PET scans, and clinical evaluations. CNNs effectively capture spatial and temporal patterns within each data modality, preserving inter-modal linkages. The proposed RNN-DBN architecture, by skillfully leveraging temporal dependencies, improves model interpretability and provides a clearer understanding of the progression of Parkinson's disease symptoms. Evaluation across diverse PD datasets demonstrates superior predictive performance compared to existing methods. This multimodal deep learning framework holds the potential to revolutionize PD diagnosis and monitoring, offering physicians a valuable tool for assessing the condition's severity. The integration of various data sources enhances the model's accuracy, providing a holistic perspective on Parkinson's disease progression. This, in turn, facilitates improved clinical decision-making and patient care. Notably, the implementation in Python achieves a remarkable accuracy of 94.87%, surpassing existing methods like EOFSC and CNN by 1.44%.

Author 1: Shaikh Abdul Hannan

Keywords: Parkinson's Disease (PD); Convolutional Neural Networks (CNN); Deep Belief Networks (DBN); Rat Swarm Optimization (RSO)

Paper 51: Efficiency of Hybrid Decision Tree Algorithms in Evaluating the Academic Performance of Students

Abstract: Educational institutions are anticipated to take substantial and proactive roles in guaranteeing students' successful program completion. Academic performance is conventionally employed to categorize and forecast students' future ability to confront post-graduation challenges. A student's academic accomplishments are instrumental in shaping exceptional individuals who may become future leaders. Using algorithms to assess and predict academic performance is a well-established practice in machine learning, encompassing techniques such as neural networks(NN), logistic regression(LR), decision trees(DT), and others. The goal of this project is to improve decision trees' ability to predict students' academic achievement via the use of data mining methods and meta-heuristic algorithms. ‎‎‎Educational data mining involves the utilization of data analysis methodologies and tools to examine the extensive data generated within educational establishments as a result of students' interactions and activities throughout their academic journey. Pelican Optimization Algorithm (POA) and Runge Kutta optimization (RKO) are utilized algorithms in developing hybrid models, both of which can efficiently search for optimal or near-optimal splits by fine-tuning the hyperparameters of decision tree models. Students' final grades were predicted through training and testing models and categorized into four classes: Excellent, Good, Acceptable, and Poor. The classification capability of a single model and optimized counterparts was evaluated using Accuracy, Recall, Precision, and F1-score in separate phases for each category. Obtained results for all models revealed that POA and RKO developed Accuracy of DTC by 1.86% and 0.87%. Also, Precision and Recall metric analysis further manifest the superiority of DTPO. Prediction based on classifiers, especially workable optimized versions such as DTPO, paves the way for institutions to raise student success rates.

Author 1: Yanxin Xie

Keywords: Academic performance; decision tree; pelican optimization algorithm; runge kutta optimization

Paper 52: A Novel Robust Stacked Broad Learning System for Noisy Data Regression

Abstract: Robust broad learning system (RBLS) demonstrates the generalization and robustness for solving uncertain data regression tasks. To enhance representation ability of RBLS, this paper aims at developing a novel robust stacked broad learning system for solving noisy data regression problems, termed as RSBLS. In our work, we expand traditional BLS into a stacked broad learning system model with deep structure of feature nodes and enhancement nodes. Furthermore, ℓ1 norm loss function is employed to update the objective function of RSBLS for processing noisy data, we apply augmented Lagrange multiplier (ALM) to get the output weights of RSBLS which keeps the effectiveness and efficiency compared with weighted loss function. Simulation results over some regression datasets with outliers demonstrate that, the proposed RSBLS performs favorably with better robustness with respect to RVFL, BLS, Huber-WBLS, KDE-WBLS and RBLS.

Author 1: Kai Zheng

Author 2: Jie Liu

Keywords: Robust; stacking; broad learning system; deep learning; neural networks

Paper 53: Deep Learning Augmented with SMOTE for Timely Alzheimer's Disease Detection in MRI Images

Abstract: Timely diagnosis of Alzheimer's Disease (AD) is pivotal for effective intervention and improved patient outcomes, utilizing Magnetic Resonance Imaging (MRI) to unveil structural brain changes associated with the disorder. This research presents an integrated methodology for early detection of Alzheimer's Disease from Magnetic Resonance Imaging, combining advanced techniques. The framework initiates with Convolutional Neural Networks (CNNs) for intricate feature extraction from structural MRI data indicative of Alzheimer's Disease. To address class imbalance in medical datasets, Synthetic Minority Over-sampling Technique (SMOTE) ensures a balanced representation of Alzheimer's Disease and non- Alzheimer's Disease instances. The classification phase employs Spider Monkey Optimization (SMO) to optimize model parameters, enhancing precision and sensitivity in Alzheimer's Disease diagnosis. This work aims to provide a comprehensive approach, improving accuracy and tackling imbalanced datasets challenges in early Alzheimer's detection. Experimental outcomes demonstrate the proposed approach outperforming conventional techniques in terms of classification accuracy, sensitivity, and specificity. With a notable 91% classification accuracy, particularly significant in medical diagnostics, this method holds promise for practical application in clinical settings, showcasing robustness and potential for enhancing patient outcomes in early-stage Alzheimer's diagnosis. The implementation is conducted in Python.

Author 1: P Gayathri

Author 2: N. Geetha

Author 3: M. Sridhar

Author 4: Ramu Kuchipudi

Author 5: K. Suresh Babu

Author 6: Lakshmana Phaneendra Maguluri

Author 7: B Kiran Bala

Keywords: Alzheimer's disease; MRI scans; Convolutional Neural Networks (CNNs); Synthetic Minority Over-sampling Technique (SMOTE); Spider Monkey Optimization (SMO)

Paper 54: Load Balancing in DCN Servers Through Software Defined Network Machine Learning

Abstract: In this research paper, we delve into the innovative realm of optimizing load balancing in Data Center Networks (DCNs) by leveraging the capabilities of Software-Defined Networking (SDN) and machine learning algorithms. Traditional DCN architectures face significant challenges in handling unpredictable traffic patterns, leading to bottlenecks, network congestion, and suboptimal utilization of resources. Our study proposes a novel framework that integrates the flexibility and programmability of SDN with the predictive and analytical prowess of machine learning. We employed a multi-layered methodology, initially constructing a virtualized environment to simulate real-world DCN traffic scenarios, followed by the implementation of SDN controllers to instill adaptiveness and programmability. Subsequently, we integrated machine learning models, training them on a substantial dataset encompassing diverse traffic patterns and network conditions. The crux of our approach was the application of these trained models to anticipate network congestion and dynamically adjust traffic flows, ensuring efficient load distribution among servers. A comparative analysis was conducted against prevailing load balancing methods, revealing our model's superiority in terms of latency reduction, enhanced throughput, and improved resource allocation. Furthermore, our research illuminates the potential for machine learning's self-learning mechanism to foresee and adapt to future network states or exigencies, marking a significant advancement from reactive to proactive network management. This convergence of SDN and machine learning, as demonstrated, ushers in a new era of intelligent, scalable, and highly reliable DCNs, demanding further exploration and investment for future-ready data centers.

Author 1: Gulbakhram Beissenova

Author 2: Aziza Zhidebayeva

Author 3: Zhadyra Kopzhassarova

Author 4: Pernekul Kozhabekova

Author 5: Bayan Myrzakhmetova

Author 6: Mukhtar Kerimbekov

Author 7: Dinara Ussipbekova

Author 8: Nabi Yeshenkozhaev

Keywords: Software defined network; DCN; machine learning; deep learning; server; load balancing; software

Paper 55: An Intelligent Fuzzy-PID Controller for Supporting Comfort Microclimate in Smart Homes

Abstract: Addressing the challenge of ensuring a comfortable indoor environment in both commercial and residential buildings through the use of heating, ventilation, and air conditioning (HVAC) systems is a critical issue. This challenge is intricately connected to the development of sophisticated multi-channel controllers to regulate temperature and humidity effectively. This academic discussion initially focuses on the development and examination of a complex, interactive, nonlinear mathematical model that encapsulates the ideal parameters for temperature and humidity to achieve the desired comfort levels. The paper then progresses to explore various methodologies in the design of temperature and humidity control systems. It delves into the traditional Proportional-Integral-Derivative (PID) controllers, a mainstay in the industry, and extends to more advanced iterations. These include the integration of PID controllers with distinct decoupled controllers and the innovative combination of PID controllers with self-adjusting parameters, which are informed by the principles of fuzzy logic. This combination is particularly significant for the processes of heating and humidification. Subsequently, the paper presents the results obtained from simulations conducted on a proposed fuzzy-PID controller using Matlab, a widely used computational tool. These simulations are crucial in evaluating the efficacy of the controller design. Additionally, the paper offers an analysis of experimental data collected over a six-month period. This data is instrumental in assessing the real-world performance of the proposed system, providing valuable insights into its practical applicability and effectiveness in managing indoor climate conditions. In summary, this comprehensive study not only lays the groundwork for an interactive model for climate control but also compares various controller designs, culminating in the proposal and evaluation of an advanced fuzzy-PID controller. This work stands as a significant contribution to the ongoing efforts to enhance indoor climate control in buildings.

Author 1: Nazbek Katayev

Author 2: Ainur Zhakish

Author 3: Nurlan Kulmyrzayev

Author 4: Assylzat Abuova

Author 5: Sveta Toxanova

Author 6: Aiymkhan Ostayeva

Author 7: Gulsim Dossanova

Keywords: HVAC; Fuzzy logic; energy management; comfort management; smart home

Paper 56: Efficient Compression for Remote Sensing: Multispectral Transform and Deep Recurrent Neural Networks for Lossless Hyper-Spectral Imagine

Abstract: Remote sensing technologies, which are essential for everything from environmental monitoring to disaster relief, enable large-scale multispectral data collection. In the field of hyper-spectral imaging, where high-dimensional data is required for precise analysis, effective compression techniques are critical for transmission and storage. In the field of hyper-spectral imaging, the development of efficient compression techniques is critical because datasets containing high-dimensional information must be transmitted and stored efficiently without sacrificing analytical precision. The paper presents advanced compression techniques that combine deep Recurrent Neural Networks (RNNs) with multispectral transforms to achieve lossless compression in hyper-spectral imaging. The Discrete Wavelet Transform (DWT) is used to efficiently capture spectral and spatial information by utilizing the properties of multispectral transforms. Simultaneously, deep RNNs are used to model the hyper-spectral data with complex dependencies, allowing for sequential compression. The overall compression efficiency that is increased by the integration of spatial and spectral information allows for reduced storage requirements and improved transmission efficiency. Python software is used to implement the proposed model. When compared to Liner Spectral Mixture Analysis (LSMA) based compression, Spatial Orientation Tree Wavelet (STW)-Wavelet Difference Reduction (WDR), and DPCM, the proposed DWT-RNN-LSTM method has a better PSNR value of 45 dB and a lower MSE of 7.50%. Adaptive compression methods are presented in order to dynamically adapt to various data properties and ensure application in various hyperspectral scenes. Studies on hyper-spectral images of various sizes and resolutions demonstrate the approach's scalability and generalization, as well as the utility and adaptability of the proposed compression framework in a variety of remote sensing scenarios.

Author 1: D. Anuradha

Author 2: Gillala Chandra Sekhar

Author 3: Annapurna Mishra

Author 4: Puneet Thapar

Author 5: Yousef A.Baker El-Ebiary

Author 6: Maganti Syamala

Keywords: Multi-Spectral transform; lossless compression; hyper-spectral data; deep recurrent neural network; compression algorithms

Paper 57: MR-FNC: A Fake News Classification Model to Mitigate Racism

Abstract: One of the most challenging tasks while processing natural language text is to authenticate the correctness of the provided information particularly for classification of fake news. Fake news is a growing source of apprehension in recent times for hate speech as well. For instance, the followers of various beliefs face constant discrimination and receive negative perspectives directed at them. Fake news is one of the most prominent reasons for various kinds of racism and stands at par with individual, interpersonal, and structural racism types observed worldwide yet it does not get much importance and remains to be neglected. In this paper, to mitigate racism, we address the fake news regarding beliefs related to Islam as a case study. Though fake news remained to be a concerning factor since the beginning of Islam, a significant increase has been noticed in it for the last three years. Additionally, the accessibility of social media platforms and the growth in their use have helped to propagate misinformation, hate speech, and unfavorable views about Islam. Based on these deductions, this study intends to categorize such anti-Islamic content and misinformation found in Twitter posts. Several preprocessing and data enhancement steps were employed on retrieved data. Word2vec and GloVe were implemented to derive deep features while TF-IDF and BOW were applied to derive textual features from the data respectively. Finally, the classification phase was performed using four Machine-based predictive analysis (ML) algorithms Random Forest (RF), Naïve Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM), and a custom deep CNN. The results when compared with certain performance evaluation measures show that on average, ML-models perform better than the CNN for the utilized dataset.

Author 1: Muhammad Kamran

Author 2: Ahmad S. Alghamdi

Author 3: Ammar Saeed

Author 4: Faisal S. Alsubaei

Keywords: Machine learning; deep learning; fake news detection; social media

Paper 58: i-Tech: Empowering Educators to Bring Experimental Learning to Classrooms

Abstract: The integration of technology in education has gained significant attention, with Virtual Reality (VR), Augmented Reality (AR), and 360° VR emerging as transformative tools for enhancing student learning experiences. Despite their potential benefits, these immersive technologies have not achieved widespread adoption in education. Educators face numerous challenges in finding suitable 360° content for their courses and integrating complex content creation tools. Creating educational 360° content often involves hiring programmers or mastering intricate programming techniques, which can be time-consuming and daunting. Educators also struggle with finding platforms to host, edit, segment video content according to topics, and add subtitles and translations to their 360° videos. To address these challenges, this paper presents the implementation and evaluation of a user-friendly prototype tool with a step-by-step graphical user interface. This high-fidelity prototype assists educators in uploading 360° content, segmenting it into chapters or topics, incorporating questions or requirements within video segments, adding subtitles and translations, and facilitating content sharing among educators. This design aims to assist teachers in publishing their 360° content while reducing the complex VR programming for them. It enables them to integrate immersive learning in their classrooms with ease. The final goal is to promote greater adoption of 360° VR content in education and enhance learning outcomes.

Author 1: Amani Alqarni

Author 2: Jieyu Wang

Author 3: Abdullah Abuhussein

Keywords: Virtual reality; 360° video; user behavior analysis; content delivery; immersive media; education; technology in education; instructional design; human-computer interaction

Paper 59: A Lightweight Neural Network for Accurate Rice Panicle Detection and Counting in Field Conditions

Abstract: Monitoring rice spikelet yield is crucial for ensuring food security, but manual observations are tedious and subjective. Deep learning approaches for automated counting often require high device resources, limiting their applicability on low-cost edge devices. This paper presents the Rice Lightweight Feature Detection Network (RLFDNet). RLFDNet designed for the field of computer vision, features a lightweight encoder and decoder, effectively decoding shallow and deep information within its neural network architecture. Innovative designs including dense feature pyramid network, reinforcement learning guidance, attention mechanisms, dynamic receptive field adjustment, and shape feature fusion enable outstanding performance in object detection and counting, even with low-resolution images. Across different elevations, ranging from 7m to 20m, RLFDNet demonstrates significantly superior accuracy and inference efficiency compared to other advanced object detection methods. With a parameter count of only 4.40 million, it achieves an impressive frame rate of 80.43 FPS on a GTX1080Ti GPU, meeting real-time application requirements on inexpensive devices. RLFDNet's exceptional performance is further highlighted by an MAE of 1.86 and an R² of 0.9461, along with an average precision of [email protected] reaching 0.91. These results underscore RLFDNet's capability as a potent and reliable visual tool for agricultural practitioners, offering promising prospects for future research endeavors.

Author 1: Wenchao Xu

Author 2: Yangxu Wang

Keywords: Computer vision; deep learning; lightweight; neural network architecture; remote sensing

Paper 60: Integrating Taguchi Method and Support Vector Machine for Enhanced Surface Roughness Modeling and Optimization

Abstract: End milling process is widely used in various industrial applications, including health, aerospace and manufacturing industries. Over the years, machine technology of end milling has grown exponentially to attain the needs of various fields especially in manufacturing industry. The main concern of manufacturing industry is to obtain good quality products. The machined products quality is commonly correlated with the value of surface roughness (Ra), representing vital aspect that can influence overall machining performance. However, finding the optimal value of surface roughness is remain as a challenging task because it involves a lot of considerations on the cutting process especially the selection of suitable machining parameters and also cutting materials and workpiece. Hence, this study presents a support vector machine (SVM) prediction model to obtain the minimum Ra for end milling machining process. The prediction model was developed with three input parameters, namely feed rate, depth of cut and spindle speed, while Ra is the output parameter. The data of end milling is collected from the case studies based on the machining experimental with titanium alloy, workpiece and three types of cutting tools, namely uncoated carbide WC-Co (uncoated), common PVD-TiAlN (TiAlN) and Supernitride coating (SNTR). The prediction result has found that SVM is an effective prediction model by giving a better Ra value compared with experimental and regression results.

Author 1: Ashanira Mat Deris

Author 2: Rozniza Ali

Author 3: Ily Amalina Ahmad Sabri

Author 4: Nurezayana Zainal

Keywords: Support Vector Machine; surface roughness; end milling; Taguchi method

Paper 61: Image Retrieval Evaluation Metric for Songket Motif

Abstract: Songket is a fine art heritage specializing in promoting the unique features of Malay identity. Past studies have shown that hundreds of Songket motifs had been produced, but unfortunately, most were not stored digitally. However, the digital collection of image data and determining its ground truth data should be given attention. This paper focuses on an evaluation metric to retrieve image Songket motifs. The initial label for each class of images in the database is ground truth data. The activity of determining the ground truth data involves two research objectives that have been discussed, namely identifying the ground truth data set of Songket Motifs involving Activity One, to obtain two ground truth data sets, precisely the training data set and the test data set involving six categories, to be specific 'Flora', 'Fauna', 'Nature', 'Cosmos', 'Food' and 'Calligraphy'. This phase test was carried out through a survey using a qualitative method, which is participatory by 15 respondents who have classified 413 specific motif images into 56 Songket motifs categories that refer to the six prominent motifs. Meanwhile, Activity Two is a validation-classification test of ground truth data sets by three experts to equate the selection of general and expert respondents to obtain training data sets for testing purposes involving six categories. After rearranging, only 50 ground truth Specific Motifs have been selected. Accordingly, the relationship coefficient correlation method is also implemented to see the relationship between two data through a statistical evaluation angle. In addition, precision and recall methods are also used to obtain precision and recall values for each ground truth data, and the F-measurement method is used to make a single evaluation. The F-measurement result for each category 'Flora': 26.7 – 100 (20 ID-Category), 'Fauna': 35.3 – 100 (6 ID-Category), 'Nature': 30.8 – 100 (5 ID-Category), 'Cosmos': 53.3 – 100 (7 ID-Category), and 'Motif': 47.6 – 100 (9 ID-Category). Using ground truth data enables image retrieval research to conduct unbiased system testing and evaluation.

Author 1: Nadiah Yusof

Author 2: Amirah Ismail

Author 3: Nazatul Aini Abd Majid

Author 4: Zurina Muda

Keywords: Heritage; songket motifs; songket motifs retrieval; ground truth data

Paper 62: An Approach to Classifying X-Ray Images of Scoliosis and Spondylolisthesis Based on Fine-Tuned Xception Model

Abstract: The vertebral column is a marvel of biological engineering and it considers a main part of the skeleton in vertebrate animals. In addition, it serves as the central axis of the human body comprising a series of interlocking vertebrae that provide structural support and flexibility. From basic works like bending and twisting to more complex actions such as walking and running, the spine's impact on human life is profound, underscoring its indispensable role in maintaining physical well-being and overall functionality. Moreover, in the hard-working schedule of people in modern life, a bunch of diseases impact on vertebral column such as spondylolisthesis and scoliosis. As a result, numerous researches were provided to take a hand in solving or avoiding these illnesses including machine learning. In this study, transfer learning and fine tuning were used for the classification of X-ray images on vertebrae sickness to avoid complex and wasted time in a medical examination process. The dataset for vertebrae illnesses X-ray images was collected at King Abdullah University Hospital and Jordan University of Science and Technology in Irbid, Jordan. It comprised 338 subjects including: 79 spondylolisthesis, 188 scoliosis, and 71 normal X-ray images. With the customized layers model in Xception that is used for image classification, we received surprisingly high results including validation accuracy, test accuracy, and F1 score in three-class classifications (i.e., spondylolisthesis, scoliosis, and normal) at 99.00%, 97.86%, and 97.86%, respectively. Additionally, two-class detection also received high accuracy values (i.e., 98.86% and 99.57%). Considering various high-performance metrics in the result indicates a robust ability to identify vertebrae diseases using X-ray images. The study found that machine learning significantly raises medical examinations compared to traditional methods, offering a myriad of benefits in terms of accuracy, efficiency, and diagnostic capabilities.

Author 1: Quy Thanh Lu

Author 2: Triet Minh Nguyen

Keywords: Transfer learning; fine tuning; spondylolisthesis; scoliosis; classification; Xception

Paper 63: Toward Enhanced Customer Transaction Insights: An Apriori Algorithm-based Analysis of Sales Patterns at University Industrial Corporation

Abstract: The University Industrial Corporation (CIU) at the National University of Jaen offers a range of consumable products, encompassing nectar, water, coffee, chocolate, and chocoteja. However, its sales transactions function without a systematic analysis. To address this, the study gathered and analyzed sales data from March to November 2023, aiming to identify and delineate associations among frequently co-purchased products, revealing underlying interdependencies and associations. Employing text mining methodologies, this study preprocessed and analyzed 1542 sales records using the Apriori algorithm, culminating in the extraction of 17 association rules. Among these rules, three standout associations were uncovered: the purchase of chocolate, chocoteja and water suggests a purchase of nectar; chocolate, nectar and water acquisitions correlate with chocoteja purchases; lastly chocolate and nectar purchases are associated with chocoteja acquisitions. These findings provide insights to augment potential production adjustments within the CIU, enabling the leveraging of established associations to boost sales and revenue. Moreover, the identified rules serve as a cornerstone for decision-makers, actionable guidance for stakeholders, enabling the identification of co-purchased products, fostering informed production planning, fine-tuning marketing strategies for customer relationship management (CRM), and enhancing CIU's market competitiveness and profitability.

Author 1: Alex Alfredo Huaman Llanos

Author 2: Lenin Quiñones Huatangari

Author 3: Jeimis Royler Yalta Meza

Author 4: Alexander Huaman Monteza

Author 5: Orestes Daniel Adrianzen Guerrero

Author 6: John Smith Rodriguez Estacio

Keywords: Apriori algorithm; association rules; Customer Relationship Management (CRM); decision making; text mining

Paper 64: Automated Detection of Autism Spectrum Disorder Symptoms using Text Mining and Machine Learning for Early Diagnosis

Abstract: Autism spectrum disorder (ASD) is a neurological condition whose etiology is still insufficiently understood. The heterogeneity of manifestations makes the diagnosis process difficult. Thus, many children are diagnosed too late, which leads to the loss of precious time that can be used for therapy. A viable solution could be to equip medical staff with modern technologies to detect autism in its early stages. The objective of this research was to investigate, through empirical means, how text mining and machine learning (ML) algorithms can aid in the early ASD diagnosis by identifying patterns and ASD symptoms in text data regarding children’s behavior that concerned parents provided. The research involved the design of an innovative technical solution based on text mining for the identification of ASD symptoms in unstructured text data describing children’s behavior and the practical implementation of the solution using Rapid Miner. The dataset was created through a controlled experiment with 44 participants, parents of children diagnosed with ASD, who answered questions about their children’s (35 boys and 9 girls) behavior. Analysis of the performance of models trained with ML algorithms: Naïve Bayes, K-Nearest Neighbors, Deep Learning and Random Forest revealed that the K-Nearest Neighbors classifier outperformed the other methods, achieving the highest accuracy of 78.69%. Results obtained using text mining and ML demonstrated the feasibility of using parents’ narratives to develop predictive models for autism symptoms detection. The achieved accuracy highlights the potential of text mining as an autonomous and time- and cost-effective method for early identification of ASD in children.

Author 1: Mihaela Chistol

Author 2: Mirela Danubianu

Keywords: Text mining; machine learning; artificial intelligence; assistive technologies; Autism Spectrum Disorder; early diagnosis; screening

Paper 65: A Novel Inter Patient ECG Arrhythmia Classification Approach with Deep Feature Extraction and 1D Convolutional Neural Network

Abstract: The World Health Organization (WHO) sheds light on the escalating prevalence of heart diseases, foreseeing a substantial rise in the years ahead, impacting a vast global population. Swift and accurate early detection becomes pivotal in managing severe complications, underscoring the urgency of timely identification. While Ventricular Ectopic Beats (V) might initially be considered normal, their frequent occurrence could serve as a potential red flag for progressing to severe conditions like atrial fibrillation, Ventricular Tachycardia, and even cardiac arrest. This accentuates the need for developing an automated approach for early detection of cardiovascular diseases (CVD). This paper presents a novel method to classify arrhythmias. Leveraging the Wavelet Scattering Transform (WST) to extract morphological features from Electrocardiogram heartbeats (ECG), these features seamlessly integrate into a 1D Convolutional Neural Network (CNN). The CNN is finely tuned to distinguish between V, Supraventricular Ectopic Beats (S), and Non-Ectopic Beats (N). Our model's performance surpasses state-of-the-art models, boasting precision, sensitivity, and specificity of 94.56%, 97.26%, and 99.54% for V, and 99.25%, 98.65%, and 93.26% for N. Remarkably, it achieves 68.01% precision, 77.75% sensitivity, and 99.14% specificity for S.

Author 1: Mohamed Elmehdi Ait Bourkha

Author 2: Anas Hatim

Author 3: Dounia Nasir

Author 4: Said El Beid

Author 5: Assia Sayed Tahiri

Keywords: Electrocardiogram (ECG); Cardiovascular Diseases (CVD); Wavelet Scattering Transform (WST); Convolutional Neural Network (CNN)

Paper 66: Leveraging Machine Learning for Enhanced Cyber Attack Detection and Defence in Big Data Management and Process Mining

Abstract: The rapidly developing field of "Commercial Operation Divergence Analysis," this research seeks to identify and understand differences in commercial systems that exceed expected results. Approaches in this domain aim to identify the characteristics of process implementations that are associated with changes in process effectiveness. This entails identifying the features of procedural behaviours that result in unpleasant results and figuring out which behaviours have the biggest impact on increased efficiency. As the scale and complexity of big data management and process mining continue to expand, the threat of cyber-attacks poses a critical challenge. This research leverages machine learning techniques for the detection and defence against cyber threats within the realm of big data management and process mining. The study introduces novel metrics such as Skewness, Coefficient of Variation, Standard Deviation, Maximum, Minimum, and Mean for assessing the security state, utilizing variables like SPI, SPEI, and SSI. The research addresses prior issues in cyber-attack detection by integrating machine learning into the specific context of big data and process mining. The novelty lies in the application of Skewness and other statistical metrics to enhance the precision of threat detection. The results demonstrate the effectiveness of the proposed methodology, showcasing promising outcomes in identifying and mitigating cyber threats in the given dataset and which makes use of Support Vector Regression (SVR), has a standard deviation of 0.9, which is consistent with the variability shown in SVM. The results demonstrate a significant achievement, with a Mean Absolute Error (MAE) of 0.98, indicating the efficacy of the proposed approach in providing accurate and timely insights for cyberattack detection and defense, thereby enhancing the overall security posture in data-intensive systems. The results highlight how well the proposed method extracts significant insights from complicated event data, with important ramifications for real-world application and decision-making procedures.

Author 1: Taviti Naidu Gongada

Author 2: Amit Agnihotri

Author 3: Kathari Santosh

Author 4: Vijayalakshmi Ponnuswamy

Author 5: Narendran S

Author 6: Tripti Sharma

Author 7: Yousef A.Baker El-Ebiary

Keywords: Machine learning; data mining; cyber-attack detection; big data; support vector regression

Paper 67: Utilizing Federated Learning for Enhanced Real-Time Traffic Prediction in Smart Urban Environments

Abstract: Federated Learning (FL), a crucial advancement in smart city technology, combines real-time traffic predictions with the potential to enhance urban mobility. This paper suggests a novel approach to real-time traffic prediction in smart cities: a hybrid Convolutional Neural Network-Recurrent Neural Network (CNN-RNN) architecture. The investigation started with the systematic collection and preprocessing of a low-resolution dataset (1.6 GB) derived from real-time Closed Circuit Television (CCTV) traffic camera images at significant intersections in Guntur and Vijayawada. The dataset has been cleaned up utilizing min-max normalization to facilitate use. The primary contribution of this study is the hybrid architecture that it develops by fusing RNN to detect temporal dynamics with CNN for geographic extraction of characteristics. While the RNN's recurrent interactions preserve hidden states for sequential processing, the CNN efficiently retrieves high-level spatial information from static traffic images. Weight adjustments and backpropagation are used in the training of the proposed hybrid model in order to enhance real-time predictions that aid in traffic management. Notably, the implementation is done with Python software. The model reaches a testing accuracy of 99.8% by the 100th epoch, demonstrating excellent performance in the results and discussion section. The Mean Absolute Error (MAE) results, which show a 4.5% improvement over existing methods like Long Short Term Memory (LSTM), Support Vector Machine (SVM), Sparse Auto Encoder (SAE), and Gated Recurrent Unit (GRU), illustrate the efficacy of the model. This demonstrates how well complex patterns may be represented by the model, yielding precise real-time traffic predictions in crowded metropolitan settings. A new era of more precise and effective real-time traffic forecasts is about to begin, thanks to the hybrid CNN-RNN architecture, which is validated by the combined strengths of FL, CNN, and RNN as well as the overall outcomes.

Author 1: Mamta Kumari

Author 2: Zoirov Ulmas

Author 3: Suseendra R

Author 4: Janjhyam Venkata Naga Ramesh

Author 5: Yousef A. Baker El-Ebiary

Keywords: Federated Learning; smart city; convolutional neural network; recurrent neural network; traffic prediction

Paper 68: Elevating Smart Industry Security: An Advanced IoT-Integrated Framework for Detecting Suspicious Activities using ELM and LSTM Networks

Abstract: The proliferation of Internet of Things (IoT) devices in smart industrial contexts necessitates robust security measures to thwart potential threats. This study addresses the escalating security challenges arising from the widespread deployment of IoT devices in smart industrial environments. Focusing on the identification and categorization of potentially harmful activities, our research introduces an innovative framework that seamlessly integrates networks of Extreme Learning Machines (ELM) with Long Short-Term Memory (LSTM). The primary goal is to significantly enhance the accuracy and efficiency of real-time detection of suspicious activities. Implemented using Python, the framework exhibits a remarkable 97.5% improvement in recognizing and accurately categorizing suspicious activities compared to traditional methods such as Conv 1D and 3D CNN. Rigorous testing on a substantial real-world dataset simulating smart industry scenarios underlines this substantial improvement over conventional approaches in identifying and precisely classifying questionable activities. The design excels in comprehending complex behavioral trends within the dynamic IoT data environment, leveraging the temporal memory retention capacity of LSTM networks. This research lays the groundwork for fortifying cybersecurity in smart industries against emerging online threats and malicious actions. The proposed framework capitalizes on the synergies between LSTM and ELM networks to achieve heightened accuracy in identifying suspicious activities, providing comprehensive and dynamic insights from real-time IoT data. These insights are crucial for proactive threat detection and prevention in smart industrial settings, contributing to an elevated level of security against evolving threats.

Author 1: Mohammad Eid Alzahrani

Keywords: Internet of Things (IoT); Smart Industries; Extreme Learning Machine (ELM); Long Short-Term Memory (LSTM); Activity Recognition

Paper 69: Enhancing Agricultural Yield Forecasting with Deep Convolutional Generative Adversarial Networks and Satellite Data

Abstract: Ensuring food security amidst growing global population and environmental changes is imperative. This research introduces a pioneering approach that integrates cutting-edge deep learning techniques. Deep Convolutional Generative Adversarial Networks (DCGANs) and Convolutional Neural Networks (CNNs) with high-resolution satellite imagery to optimize agricultural yield prediction. The model leverages DCGANs to generate synthetic satellite images resembling real agricultural settings, enriching the dataset for training a CNN-based yield estimation model alongside actual satellite data. DCGANs facilitate data augmentation, enhancing the model's generalization across diverse environmental and seasonal scenarios. Extensive experiments with multi-temporal and multi-spectral satellite image datasets validate the proposed method's effectiveness. Trained CNN adeptly discerns intricate patterns related to crop growth phases, health, and yield potential. Leveraging Python software, the study confirms that integrating DCGANs significantly enhances agricultural production forecasting compared to conventional CNN-based approaches. Against established optimization methods like RCNN, YOLOv3, Deep CNN, and Two Stage Neural Networks, the proposed DCGAN-CNN fusion achieves 98.6% accuracy, a 3.62% improvement. Synthetic images augment model resilience by exposing it to varied situations and enhancing adaptability to diverse geographic regions and climatic shifts. Moreover, the research delves into CNN model interpretability, elucidating learnt features and their correlation with yield-related factors. This paradigm promises to advance agricultural output projections, advocate sustainable farming, and aid policymakers in addressing global food security amidst evolving environmental challenges.

Author 1: D. Anuradha

Author 2: Ramu Kuchipudi

Author 3: B Ashreetha

Author 4: Janjhyam Venkata Naga Ramesh

Author 5: Ayadi Rami

Keywords: Agricultural yield prediction; DCGANs; CNN; satellite imagery; data augmentation; synthetic image generation

Paper 70: Analyzing Multiple Data Sources for Suicide Risk Detection: A Deep Learning Hybrid Approach

Abstract: In the current digital landscape, social media’s extensive user-generated content presents a unique opportunity for identifying emotional distress signals. With suicide rates on the rise, this study takes aid of Natural Language Processing (NLP) and Sentiment Analysis to detect suicide risk. Centering primarily around deep learning (DL) architectures, including Convolutional Neural Network (CNN), Bidirectional Gated Recurrent Unit (Bi-GRU) and their combined hybrid BiGRU-CNN model, the research incorporates machine learning (ML) for comparative analysis through multisource datasets from Reddit and Twitter. The methodology commenced with data pre-processing, followed by exploring word embedding techniques. This research included an analysis of both Word2Vec variants as well as pretrained GloVe embeddings, where Skip-Gram paired with Adam optimizer showed superior results. For thorough evaluation, Receiver Operating Characteristic (ROC) curves, Confusion Matrix and Accuracy-Loss graphs were utilized. Furthermore, generalizability of employed models was testified and evaluated by in-depth inspections. The process was accomplished by activating manual input test, cross dataset test and k-fold cross validation procedures. In the course of scrutinizing, the proposed BiGRU-CNN model outperformed the traditional DL and ML models with consistent and reliable performance. Correspondingly, the proposed model achieved accuracies of 93.07% and 92.47% on the respective datasets which advocate its potential as a tool for the early detection of suicidal thought.

Author 1: Saraf Anika

Author 2: Swarup Dewanjee

Author 3: Sidratul Muntaha

Keywords: BiGRU-CNN hybrid; multisource dataset; word embeddings; NLP; sentiment analysis; cross-dataset testing

Paper 71: Enhancing the Odia Handwritten Character and Numeral Recognition System's Performance with an Ensemble of Deep Neural Networks

Abstract: Offline handwritten character recognition (OHCR) is considered a challenging task in pattern recognition due to the inter-class similarity and intra-class variations among the symbols present in the alphabet set. In this work, a learning-based weighted average ensemble of deep neural network models (WEnDNN) is proposed to classify the 10 digits and 47 characters present in the alphabet set of Odia language, an official language of India. To build the base model for the ensemble network (EnDNN), three suitable convolutional neural networks (CNN), are designed and trained from scratch. The WEnDNN's accuracy is increased by using a grid search approach to determine the ideal weight allocations to give to the top-performing model. The performance of the WEnDNN model is compared with several standard machine learning models, which take the non-handcrafted features extracted from the finely tuned, pre-trained VGG16 model and a combination of Gabor and pixel intensity values to create handcrafted features. On several benchmark handwritten datasets, including NITR Odia characters (OHCS v1.0), ISI Kolkata Odia numerals, and IITBBS Odia numerals, the performance of the proposed WEnDNN model is assessed and compared. The experimental results demonstrate that, in terms of recognition accuracy, the proposed approach beats other state-of-the-art approaches.

Author 1: Mamatarani Das

Author 2: Mrutyunjaya Panda

Author 3: Soumya Sahoo

Keywords: Odia language; ensemble learning; machine learning; Gabor features; CNN; DNN

Paper 72: Monitoring Student Attendance Through Vision Transformer-based Iris Recognition

Abstract: In the context of the ongoing digital transformation, the effective monitoring of student attendance holds paramount significance for educational establishments. This study presents an innovative approach using Vision Transformer technology for iris recognition to automate student attendance tracking. We fine-tuned Vision Transformer models, specifically ViT-B16, ViT-B32, ViT-L16, and ViT-L32, using the CASIA-Iris-Syn dataset and focused on overcoming challenges related to intra-class variation through data augmentation techniques, including rotation, shearing, and brightness adjustments. The results reveal that ViT-L16 is the most proficient, achieving an impressive accuracy of 95.69%. Comparative analysis with prior methodologies, specifically those employing Vision Transformer with Convolutional Neural Network, underscores the superiority of our proposed ViT-L16 model. This superiority is evident across various metrics, including accuracy, precision, recall, and F1 score. The experimental setup involves the use of Jupyter Notebook, Python technologies, TensorFlow, and Keras, emphasizing evaluations based on loss, accuracy, and Confusion Matrix. ViT-L16 consistently outshines other models, showcasing its resilience in iris recognition for student attendance. This research marks a significant step towards modernizing attendance systems, offering an accurate and automated solution suitable for the evolving needs of educational settings. Future work could explore integrating additional biometric modalities and refining Vision Transformer architecture for enhanced performance and broader application in educational environments.

Author 1: Slimane Ennajar

Author 2: Walid Bouarifi

Keywords: Iris Recognition; Vision transformer; student attendance; vision transformer models; educational technology

Paper 73: Employing a Hybrid Convolutional Neural Network and Extreme Learning Machine for Precision Liver Disease Forecasting

Abstract: This paper discusses the critical relevance of precise forecasting in liver disease, as well as the need for early identification and categorization for immediate action and personalized treatment strategies. The paper describes a unique strategy for improving liver disease classification using ultrasound image processing. The recommended technique combines the properties of the Extreme Learning Machine (ELM), Convolutional Neural Network (CNN), along Grey Wolf Optimisation (GWO) to form an integrated model known as CNN-ELM-GWO. The data is provided by Pakistan's Multan Institute of Nuclear Medicine and Radiotherapy, and it is then pre-processed utilizing bilateral and optimal wavelet filtering techniques to increase the dataset's quality. To properly extract significant visual information, feature extraction employs a deep CNN architecture using six convolutional layers, batch normalization, and max-pooling. The ELM serves as a classifier, whereas the CNN is a feature extractor. The GWO algorithm, based on grey wolf searching strategies, refines the CNN and ELM hyperparameters in two stages, progressively boosting the system's classification accuracy. When implemented in Python, CNN-ELM-GWO exceeds traditional machine learning algorithms (MLP, RF, KNN, and NB) in terms of accuracy, precision, recall, and F1-score metrics. The proposed technique achieves an impressive 99.7% accuracy, revealing its potential to significantly enhance the classification of liver disease by employing ultrasound images. The CNN-ELM-GWO technique outperforms conventional approaches in liver disease forecasting by a substantial margin of 27.5%, showing its potential to revolutionize medical imaging and prospects.

Author 1: Araddhana Arvind Deshmukh

Author 2: R. V. V. Krishna

Author 3: Rahama Salman

Author 4: S Sandhiya

Author 5: Balajee J

Author 6: Daniel Pilli

Keywords: Liver disease prognosis; convolutional neural network extreme learning machine; grey wolf optimization; patient care

Paper 74: Personalized Recommendation Algorithm Based on Trajectory Mining Model in Intelligent Travel Route Planning

Abstract: With the increasing demand for personalized travel, traditional travel route planning methods are no longer able to meet the diverse needs of users. In view of this, on the ground of the analysis of user trajectory data at the temporal and spatial levels, a new scenic spot recommendation model is proposed by combining personalized recommendation algorithms. Meanwhile, improved genetic algorithm and minimum spanning tree algorithm were introduced to adjust the structure of the personalized recommendation model. After matching the visit sequence of scenic spots, the final new personalized tourism route recommendation model was proposed. The experiment demonstrates that the optimal pause time for the personalized scenic spot recommendation model is 45 minutes, the pause distance is 15 meters, and the clustering radius is 500 meters. And the model has the highest accuracy in the Tok-10 testing environment, with a maximum value of 90%. In addition, the new personalized tourism route recommendation model has the highest accuracy of 85.6%, the highest recall rate of 88.7%, the highest F1 value of 92.4%, and an average convergence rate of 88.9%. In summary, the new scenic spot and route recommendation model proposed in the study can achieve more intelligent and personalized travel route planning, providing new guidance for the intelligent development of travel route recommendation.

Author 1: Jingya Shi

Author 2: Qianyao Sun

Keywords: Trajectory mining; personalized recommendations; travel routes; genetic algorithm; visiting sequence of scenic spots

Paper 75: Animation Media Art Teaching Design Based on Big Data Fusion Technology

Abstract: Animation, as an ancient art expression form, still has vigorous development, and the need for animation talents in society is increasing daily. This study first introduces the definition of animation and the development of animation at home and abroad. After that, the classification regression tree algorithm's principle and function theorem are described. This study divides the data into original and new animations based on big data fusion technology. It establishes a media art teaching system with search, recommendation, and playback as the three cores. Additionally, iteration is used to calculate the optimal hidden semantic matrix, a comparison is made between the benefits and drawbacks of the Sigmoid, Tanh, and ReLU functions, and lastly, the activation function chosen is the ReLU function. Compared with the loss value in the ideal case, the experimental findings comply with the likely criteria, and the categorical regression tree algorithm model predicts an error rate that falls within acceptable limits. Practically speaking, it is known that when the hidden factor dimension is 12, the system model works best for characterizing animation features. The comparison shows that the non-standard collaborative filtering recommendation system is inferior to the recommendations filtered by the categorical regression tree algorithm model. Following the use of the system, the students' drawing and directing abilities, animation scope, and animation appreciation level all improved significantly. The questionnaire survey concluded that the teachers and students of animation majors in universities were satisfied with the system.

Author 1: Rongjuan Wang

Author 2: Yiran Tao

Keywords: Animation; big data fusion; classification regression tree algorithm; media art teaching system

Paper 76: Advancing Human Action Recognition and Medical Image Segmentation using GRU Networks with V-Net Architecture

Abstract: Human Action Recognition and Medical Image Segmentation study presents a novel framework that leverages advanced neural network architectures to improve Medical Image Segmentation and Human Action Recognition (HAR). Gated Recurrent Units (GRU) are used in the HAR domain to efficiently capture complex temporal correlations in video sequences, yielding better accuracy, precision, recall, and F1 Score than current models. In computer vision and medical imaging, the current research environment highlights the significance of advanced techniques, especially when addressing problems like computational complexity, resilience, and noise in real-world applications. Improved medical image segmentation and human action recognition (HAR) are of growing interest. While methods such as the V-Net architecture for medical picture segmentation and Spatial Temporal Graph Convolutional Networks (ST-GCNs) for HAR have shown promise, they are constrained by things like processing requirement and noise sensitivity. The suggested methods highlight the necessity of sophisticated neural network topologies and optimisation techniques for medical picture segmentation and HAR, with further study focusing on transfer learning and attention processes. A Python tool has been implemented to perform min-max normalization, utilize GRU for human action recognition, employ V-net for medical image segmentation, and optimize with the Adam optimizer, with performance evaluation metrics integrated for comprehensive analysis. This study provides an optimised GRU network strategy for Human Action Recognition with 92% accuracy, and a V-Net-based method for Medical Image Segmentation with 88% Intersection over Union and 92% Dice Coefficient.

Author 1: Dustakar Surendra Rao

Author 2: L. Koteswara Rao

Author 3: Vipparthi Bhagyaraju

Author 4: P. Rohini

Keywords: Human action recognition; medical image segmentation; grated rectifier unit; V-net architecture; neural network

Paper 77: Occupancy Measurement in Under-Actuated Zones: YOLO-based Deep Learning Approach

Abstract: The challenge of accurately detecting and identifying individuals within under-actuated zones presents a relevant research problem in occupant detection. This study aims to address the challenge of occupant detection in under-actuated zones through the utilization of the You Only Look Once version 8 (YOLO v8) object detection model. The research methodology involves a comprehensive evaluation of YOLO v8's performance across three distinct zones, where its precision, accuracy, and recall capabilities in identifying occupants are rigorously assessed. The outcomes of this performance evaluation, expressed through quantitative metrics, provide compelling evidence of the efficacy of the YOLO v8 model in the context of occupant detection in under-actuated zones. Across these three diverse under-actuated zones, YOLO v8 consistently exhibits remarkable mean Average Precision (mAP) scores, achieving 99.2% in Zone 1, 78.3% in Zone 2, and 96.2% in Zone 3. These mAP scores serve as a testament to the model's precision, indicating its proficiency in accurately localizing and identifying occupants within each zone. Furthermore, YOLO v8 demonstrates impressive efficiency in executing occupant detection tasks. The model boasts rapid processing times, with all three zones being analyzed in a matter of milliseconds. Specifically, YOLO v8 achieves execution times of 0.004 seconds in both Zone 1 and Zone 3, while Zone 2, which entails slightly more computational effort, still maintains an efficient execution time of 0.024 seconds. This efficiency constitutes a pivotal advantage of YOLO v8, as it ensures expeditious and effective occupant detection in the context of under-actuated zones.

Author 1: Ade Syahputra

Author 2: Yaddarabullah

Author 3: Mohammad Faiz Azhary

Author 4: Aedah Binti Abd Rahman

Author 5: Amna Saad

Keywords: YOLO; HVAC system; occupant’s position; occupant calculation; under-actuated zone

Paper 78: Efficient Simulation of Light Scattering Effects in the Atmosphere

Abstract: Atmospheric light scattering encompasses intricate physical process, including diverse scattering mechanisms and optical parameters. Addressing the challenges posed by the computationally intensive task of deciphering this phenomenon, this study introduces an efficient real-time simulation strategy. The proposed approach employs a physics-driven atmospheric modeling, leveraging a unified phase function to emulate both Rayleigh and Mie scattering phenomena. The scattering integral is approximated and discretized using the concept of ray-marching to solve the scattering integral. Based on the characteristics of different light sources, accurate ray-marching lengths are determined, streamlining the computational trajectory of the light path. Additionally, the introduction of texture dithering enhances the randomness of the initial sampling positions. The Shadow Map algorithm is adeptly employed to generate shadow mapping textures, eliminating the need for light calculations within shadowed regions, thereby reducing the number of samples and computational workload. Finally, color synthesis is used to determine the rendering color of the atmosphere under various fog density conditions. Experimental results show that this approach significantly improves rendering efficiency, and achieves real-time rendering while maintaining a realistic light scattering effect compared with other advanced light scattering rendering methods.

Author 1: Huiling Guo

Author 2: Xiliang Ren

Author 3: Jing Zhao

Author 4: Yong Tang

Keywords: Light scattering; ray marching; jittered sampling; color synthesis; real-time rendering

Paper 79: Semantic Information Classification of IoT Perception Data Based on Density Peak Fast Search Clustering Algorithm

Abstract: In the rapidly developing field of the Internet of Things today, effective processing and analysis of perceptual data has become crucial. The perception data of the Internet of Things is usually large, diverse, and presents high-dimensional characteristics, which poses new challenges to data clustering algorithms. This study utilizes the K-center point algorithm to optimize the density peak fast search clustering algorithm, proposes a new clustering algorithm, and applies it to the research of semantic classification of perception data in the Internet of Things. Firstly, the K-center algorithm was used to optimize the clustering center optimization process of the density peak fast search clustering algorithm. Then, the optimized algorithm was applied to the automatic semantic classification model. Thus, a new automatic semantic annotation model for IoT aware data has been established. The research results showed that the classification accuracy of the proposed optimization algorithm was as high as 0.98, and the running stability of the automatic semantic annotation model optimized using this algorithm was as high as 0.99, with a running time as low as 1s. In summary, the automatic semantic annotation model built in this study can effectively improve the efficiency and accuracy of semantic classification, thereby providing more accurate and efficient data support for intelligent services.

Author 1: Lin Chen

Author 2: Jinli Hu

Author 3: Weisheng Wang

Keywords: Clustering algorithm; Internet of Things; perceived data; classification; peak density; semantic information

Paper 80: Structure-Aware Scheduling Algorithm for Deadline-Constrained Scientific Workflows in the Cloud

Abstract: Cloud computing provides pay-per-use IT services through the Internet. Although cloud computing resources can help scientific workflow applications, several algorithms face the problem of meeting the user’s deadline while minimising the cost of workflow execution. In the cloud, selecting the appropriate type and the exact number of VMs is a major challenge for scheduling algorithms, as tasks in workflow applications are distributed very differently. Depending on workflow requirements, algorithms need to decide when to provision or de-provision VMs. Therefore, this paper presents an algorithm for effectively selecting and allocating resources. Based on the workflow structure, it decides the type and number of VMs to use and when to lease and release them. For some structures, our proposed algorithm uses the initial rented VMs to schedule all tasks of the same workflow to minimise data transfer costs. We evaluate the performance of our algorithm by simulating it with synthetic workflows derived from real scientific workflows with different structures. Our algorithm is compared with Dyna and CGA approaches in terms of meeting deadlines and execution costs. The experimental results show that the proposed algorithm met all the deadline factors of each workflow, while the CGA and Dyna algorithms met 25% and 50%, respectively, of all the deadline factors of all workflows. The results also show that the proposed algorithm provides more cost-efficient schedules than CGA and Dyna.

Author 1: Ali Al-Haboobi

Author 2: Gabor Kecskemeti

Keywords: Workflow scheduling; workflow structure; cloud computing; resource provisioning; deadline constrained; infrastructure as a service

Paper 81: Prominent Security Vulnerabilities in Cloud Computing

Abstract: This research study examines the significant security vulnerabilities and threats in cloud computing, analyzes their potential consequences for enterprises, and proposes effective solutions for mitigating these vulnerabilities. This paper discusses the increasing significance of cloud security in a time characterized by rapid data expansion and technological progress. The paper examines prevalent vulnerabilities in cloud computing, including cloud misconfigurations, data leakage, shared technology threats, and insider threats. It emphasizes the necessity of adopting a proactive and comprehensive approach to ensure cloud security. The report places significant emphasis on the shared responsibility paradigm, adherence to industry laws, and the dynamic nature of cybersecurity threats. The situation necessitates the cooperation of researchers, cybersecurity professionals, and enterprises to proactively address these difficulties. This partnership aims to provide a thorough manual for organizations aiming to bolster their cloud security measures and safeguard valuable data in an ever-evolving digital landscape.

Author 1: Alanoud Alquwayzani

Author 2: Rawabi Aldossri

Author 3: Mounir Frikha

Keywords: Cloud computing; vulnerabilities; cloud security; cloud misconfigurations; data loss; threats

Paper 82: DDoS Attacks Detection in IoV using ML-based Models with an Enhanced Feature Selection Technique

Abstract: The Internet of Vesicles (IoV) is an open and integrated network system with high reliability and security control capabilities. The system consists of vehicles, users, in-frastructure, and related networks. Despite the many advantages of IoV, it is also vulnerable to various types of attacks due to the continuous and increasing growth of cyber security attacks. One of the most significant attacks is a Distributed Denial of Service (DDoS) attack, where an intruder or a group of attackers attempts to deny legitimate users access to the service. This attack is performed by many systems, and the attacker uses high-performance processing units. The most common DDoS attacks are User Datagram Protocol (UDP) Lag and, SYN Flood. There are many solutions to deal with these attacks, but DDoS attacks require high-quality solutions. In this research, we explore how these attacks can be addressed through Machine Learning (ML) models. We proposed a method for identifying DDoS attacks using ML models, which we integrate with the CICDDoS2019 dataset that contains instances of such attacks. This approach also provides a good estimate of the model’s performance based on feature extraction strategic, while still being computationally efficient algorithms to divide the dataset into training and testing sets. The best ML models tested in the UDP Lag attack, Decision Tree (DT) and Random Forest (RF) had the best results with a precision, recall, and F1 score of 99.9%. In the SYN Flood attack, the best-tested ML models, including K-Nearest Neighbor (KNN), DT, and RF, demonstrated superior results with 99.9% precision, recall, and F1-score.

Author 1: Ohoud Ali Albishi

Author 2: Monir Abdullah

Keywords: Random forest; IoV; DDoS; feature selection

Paper 83: A Review on DDoS Attacks Classifying and Detection by ML/DL Models

Abstract: Internet security is under serious threat due to Distributed Denial of Service (DDoS) attacks. These attacks inflict considerable damage by disrupting network services, resulting in the impairment and complete disablement of system functions. The accurate classification and detection of DDoS attacks is extremely important. We provide a review of different models of Machine Learning (ML)/Deep Learning (DL)-based DDoS attack detection used by researchers that consider different classifiers. Our analysis indicates a heightened emphasis on ML-based classifiers where 22% of studies opted for the widely recognized SVM classifier. For DL-based, 27% of the studies opted for the widely recognized CNN. While the majority of researchers have formulated their datasets, NSL-KDD was employed in 55% of the studies. In addition, we discussed the future directions and challenges of DDoS detection.

Author 1: Haya Malooh Alqahtani

Author 2: Monir Abdullah

Keywords: Classification; DDoS attacks; machine learning; cybersecurity; detection

Paper 84: TPMN: Texture Prior-Aware Multi-Level Feature Fusion Network for Corrugated Cardboard Parcels Defect Detection

Abstract: Surface defect detection is the task of identifying and localizing defects on the surface of an object, which is a widely applied task in various industries. In the logistics industry, logistics companies need to monitor the condition of goods for potential defects throughout the entire logistics process for effective logistics quality control. However, effective defect detection methods are still lacking for courier packages using corrugated cardboard boxes, which rely on judging whether deformation and leakage have occurred by examining areas on their surface with abundant texture. Specifically, the defect rate and supporting structure of the packages are influenced by temperature and humidity, and the openings and bends of defects are inconsistent. This results in defective packages having rich and non-uniform texture features. Moreover, convolutional neural networks struggle to effectively extract low-level semantic texture features of defects and perceive multi-level image features of packages. Considering the above challenges, we propose a novel texture prior-aware multi-level feature fusion network (TPMN). We first introduce prior knowledge and attention mechanisms to enable the neural network to focus on extracting low-level texture features from the image in the early stages. We also design a multi-level feature fusion method to integrate features from different levels, avoiding the gradual loss of low-level semantic information in CNN and enabling comprehensive perception of multi-level image features. To support further research, we contribute the cardboard-boxes-dataset, comprising 1210 images of packages. Experiments on this dataset showcase the superior performance of TPMN, even in few-shot learning scenarios, demonstrating its effectiveness in surface defect detection within the logistics and supply chain domains.

Author 1: Xing He

Author 2: Haoxiang Fan

Author 3: Cuifeng Du

Author 4: Xingyu Zhu

Author 5: Yuyu Zhou

Author 6: Renzhang Chen

Author 7: Zhefu Li

Author 8: Guihua Zheng

Author 9: Yuansheng Zhong

Author 10: Changjiang Liu

Author 11: Jiandan Yang

Author 12: Quanlong Guan

Keywords: Logistics; surface defect detection; multi-level feature fusion; prior attention; corrugated cardboard boxes

Paper 85: An Ensemble Dynamic Model and Bio-Inspired Feature Selection Method-based Decision Support System for Predicting Multiple Organ Dysfunction Syndrome in the ICU

Abstract: Multiple Organ Dysfunction Syndrome (MODS) is one of the most common and severe conditions affecting patients admitted to intensive care units (ICUs). It is characterized by the simultaneous failure or dysfunction of at least two organ systems. Although no specific remedy for MODS has been identified to date, early diagnosis and adequate organ support can significantly improve patient outcomes. Identifying patients at risk of developing MODS in the ICU is challenging. Currently, several methods are used for this purpose, including scoring systems like SOFA and MOD Score, as well as machine learning-based approaches. However, these methods often have limitations. Some require invasive features, making them complex to use in a smart healthcare system. Others suffer from a lack of performance due to various problems, which can potentially lead to unreliable predictions. Feature selection can improve ML models’ performance. Recently, bio-inspired feature selection techniques have shown promise in improving the performance of machine learning methods in many domains, but their effectiveness in MODS prediction has not yet been evaluated. Additionally, research on early MODS prediction, particularly utilizing time-series data and dynamic ensemble methods, remains limited. To fill this gap, the present research used state-of-the-art machine learning algorithms, namely dynamic ensemble techniques, to predict patients at risk of developing MODS in the ICU. Dynamic ensembles are new methods that select an ensemble of the best-performing models for every new test case. We compared the performance of these models with full features and with feature selection. Three nature-inspired meta-heuristic optimization models, namely the binary bat algorithm (BBA), grey wolf optimization (GWO), and genetic algorithm (GA), were evaluated to select the optimal feature subset. The models were built using non-invasive patient features and time-series data from the first 12 hours of ICU admission. The results showed that feature selection significantly improved the performance of dynamic ensemble models. Notably, the METADES model, employing grey wolf optimization for feature selection, demonstrated the best performance in terms of accuracy(96.5%), F1 score (96.4%), precision (97.2%), recall (95.7%), and area under the ROC curve (AUC) (98.4%). These findings highlight the potential and effectiveness of our approach for early MODS prediction in ICUs.

Author 1: Anas Maach

Author 2: El Houssine El Mazoudi

Author 3: Jamila Elalami

Author 4: Noureddine Elalami

Keywords: Ensemble dynamic model; MODS prediction; decision support system; Bio-Inspired feature selection

Paper 86: Cephalometric Landmarks Identification Through an Object Detection-based Deep Learning Model

Abstract: In the field of orthodontics, the accurate identification of cephalometric landmarks in dental radiography plays a crucial role in ensuring precise diagnoses and efficient treatment planning. Previous studies have demonstrated the impressive capabilities of advanced deep learning models in this particular domain. However, due to the ever-changing technological landscape, it is imperative to consistently investigate and explore emerging algorithms to further improve efficiency in this field. The present study centers around the assessment of the effectiveness of YOLOv8, the most recent version of the ’You Only Look Once (YOLO)’ algorithm series, with a particular emphasis on its autonomous capability to accurately identify cephalometric landmarks. In this study, a thorough examination was con-ducted to evaluate the YOLOv8 algorithm efficiency in detecting cephalometric landmarks. The assessments encompassed various aspects such as precision, adaptability in challenging conditions, and a comparative analysis with alternative algorithms. The predefined proximities of 2mm, 2.5mm, and 3mm were utilized for the comparisons. By focusing on its potential as a noteworthy breakthrough, the investigation seeks to ascertain whether the recent enhancements indeed bring about a significant stride in the precise identification of cephalometric landmarks.

Author 1: Idriss Tafala

Author 2: Fatima-Ezzahraa Ben-Bouazza

Author 3: Aymane Edder

Author 4: Oumaima Manchadi

Author 5: Mehdi Et-Taoussi

Author 6: Bassma Jioudi

Keywords: Cephalometry; YOLOv8; landmark detection; orthodontics

Paper 87: A Robust License Plate Detection and Recognition Framework for Arabic Plates with Severe Tilt Angles

Abstract: This paper addresses the challenge of accurately detecting and recognizing Arabic license plates, particularly those subjected to severe tilt angles. It presents a robust license plate detection and recognition framework that consists three main steps: plate detection and segmentation, plate perspective correction, and vehicle number recognition. In the first step, a mask R-CNN model is used to detect the plate location, providing pixel-wise labels of identified plates’ areas. Following this, a perspective correction technique is used to obtain a clear and rectangular image of each license plate in the image. Lastly, the framework employs a Bidirectional Long Short-Term Memory (Bi-LSTM) model for accurate vehicle number recognition. The framework’s efficacy is demonstrated through its application to build a plate recognition system tailored for Egyptian license plates. The system was tested on a dataset collected from campus gate cameras at Zewail city of science and technology, achieving a character accuracy of 97%.

Author 1: Khaled Hefnawy

Author 2: Ahmed Lila

Author 3: Elsayed Hemayed

Author 4: Mohamed Elshenawy

Keywords: License plate detection; license plate recognition; feature extraction; Mask R-CNN; object detection

Paper 88: Future Iris Imaging with Advanced Fuzzified Histogram Equalization

Abstract: Images captured under low lighting frequently exhibit low brightness, low contrast, and a small grayscale. These features can affect the individual’s view and severely limit the performance of machine vision systems, particularly when data annotation is involved. Hence, the issues motivate this study to examine the effectiveness of advanced fuzzified histogram equalization for image enhancement. A comparative study was conducted based on the low lighting condition of iris images to evaluate three image enhancement methods: Advanced Fuzzified Histogram Equalization (AFHE), Contrast Limited Adaptive Histogram Equalization (CLAHE), and Fuzzy Contrast Enhancement (FCE) using the MIREIS dataset. The Gaussian membership functions (GMF) were modified accordingly to satisfy the suitable pixel intensity of the input iris images. The results were compared using the peak signal-to-noise ratio (PSNR) value, including the central processing unit (CPU) times. As a result, the AFHE showed a better PSNR value at 76.02db with faster CPU times at 4.04s compared to CLAHE and FCE. Although the PSNR value of HE is slightly lower than CLAHE (0.3%) and FCE (0.7%), AFHE improved the image’s quality and brightness, which can help other researchers with the data annotation process. The performance of the proposed methods was validated by comparing them with state-of-the-art methods. The results demonstrated that AFHE, CLAHE, and FCE exceeded other HE, AHE, CLAHE, and hybrid HE using fuzzy approaches that employed PSNR metrics.

Author 1: Nurul Amirah Mashudi

Author 2: Norulhusna Ahmad

Author 3: Rudzidatul Akmam Dziyauddin

Author 4: Norliza Mohd Noor

Keywords: Image enhancement; fuzzy logic; histogram equalization; CLAHE; iris recognition

Paper 89: IMEO: Anomaly Detection for IoT Devices using Semantic-based Correlations

Abstract: In the Internet of Things (IoT) security, anomalies due to attacks or device malfunctions can have serious consequences in our daily lives. Previous solutions have been struggling with high rates of false alarms and missing many actual anomalies. They also take a long time to detect anomalies even if they successfully detect anomalies. To overcome the limitations, this paper proposes a novel anomaly detection system, named IoT Malfunction Extraction Observer (IMEO), that utilizes semantics and correlation information for smart homes. Given IoT devices installed at home, IMEO creates virtual correlations based on semantic information such as applications, device types, relation-ships, and installation locations. The generated correlations are validated and improved using event logs extracted from smart home applications. The finally extracted correlations are then used to simulate the normal behaviors of the smart home. Any discrepancy between the actual state of a device and the simulated state is reported as abnormal while comparing correlations and event logs. IMEO also utilizes the observation that malfunctions of IoT devices occur repeatedly. An anomaly database is created and used so that repetitive malfunctions are quickly detected, which eventually reduces processing time. This paper builds a smart home testbed on a real-world residential house and deploys IoT devices. Six different types of anomalies are analyzed, synthesized, and injected to the testbed, with which IMEO’s detection performance is evaluated and compared with the state-of-the-art correlation-only detection method. Experimental results demonstrate that the proposed method achieves higher performance of detection accuracy with faster processing time.

Author 1: Seungmin Oh

Author 2: Jihye Hong

Author 3: Daeho Kim

Author 4: Eun-Kyu Lee

Author 5: Junghee Jo

Keywords: Security; anomaly detection; semantics; Internet of Things; attack

Paper 90: Cross-Modal Sentiment Analysis Based on CLIP Image-Text Attention Interaction

Abstract: Multimodal sentiment analysis is a traditional text-based sentiment analysis technique. However, the field of multi-modal sentiment analysis still faces challenges such as inconsistent cross-modal feature information, poor interaction capabilities, and insufficient feature fusion. To address these issues, this paper proposes a cross-modal sentiment model based on CLIP image-text attention interaction. The model utilizes pre-trained ResNet50 and RoBERTa to extract primary image-text features. After contrastive learning with the CLIP model, it employs a multi-head attention mechanism for cross-modal feature interaction to enhance information exchange between different modalities. Subsequently, a cross-modal gating module is used to fuse feature networks, combining features at different levels while controlling feature weights. The final output is fed into a fully connected layer for sentiment recognition. Comparative experiments are conducted on the publicly available datasets MSVA-Single and MSVA-Multiple. The experimental results demonstrate that our model achieved accuracy rates of 75.38%and 73.95% , and F1-scores of 75.21% and 73.83% on the mentioned datasets, respectively. This indicates that the proposed approach exhibits higher generalization and robustness compared to existing sentiment analysis models.

Author 1: Xintao Lu

Author 2: Yonglong Ni

Author 3: Zuohua Ding

Keywords: Multi-modal; image-text interaction; multi-head attention mechanism; sentiment analysis; cross-modal fusion

Paper 91: Automation Process for Learning Outcome Predictions

Abstract: This paper presents a comprehensive study on the evaluation of algorithms for automating learning outcome predictions, with a focus on the application of machine learning techniques. We investigate various predictive models (logistic regression, random forest, gaussian naive bayes, k-nearest neighbors and support vector regression) to assess their efficacy in forecasting student performance in educational settings. Our experimental approach involves the application of these models to predict the outcomes of a specific course, analyzing their accuracy and reliability. We also highlight the significance of an automation process in facilitating the practical application of these predictive models. This study highlights the promise of machine learning in advancing educational assessment and paves the way for further investigations into enhancing the adaptability and inclusivity of algorithms in various educational settings.

Author 1: Minh-Phuong Han

Author 2: Trung-Tung Doan

Author 3: Minh-Hoan Pham

Author 4: Trung-Tuan Nguyen

Keywords: Machine learning; predictive learning outcomes; education; logistic regression; k-nearest neighbors; Gaussian Naive Bayes; Random Forest; support vector regression

Paper 92: Enhancing K-means Clustering Results with Gradient Boosting: A Post-Processing Approach

Abstract: As the volume and complexity of data continue to grow exponentially, finding efficient and accurate clustering algorithms has become crucial for many applications. K-means clustering is a widely used unsupervised machine learning technique for data analysis and pattern recognition. Despite its popularity, k-means suffers from certain limitations, such as sensitivity to initial conditions, difficulty in determining the optimal number of clusters, and the potential for misclassification. This research paper proposes an enhanced approach for improving the accuracy and performance of the k-means clustering algorithm by incorporating post-processing techniques using a gradient boosting algorithm. The proposed method comprises training the gradient boosting model on the labeled training set, i.e., the samples with correct cluster assignments obtained from the k-means algorithm, to predict the correct cluster assignments for the misclassified samples in the testing set. This results in refined cluster assignments for the testing set. The k-means algorithm is only used initially to cluster the data and obtain initial cluster assignments. The effectiveness of the proposed approach is validated through experiments on several benchmark datasets, and the results show a significant improvement in clustering accuracy and robustness compared to the standard k-means algorithm. The proposed approach has the potential to enhance the performance of k-means in various real-world applications and domains.

Author 1: Mousa Alzakan

Author 2: Hissah Almousa

Author 3: Arwa Almarzoqi

Author 4: Mohammed Alghasham

Author 5: Munirah Aldawsari

Author 6: Mohammed Al-Hagery

Keywords: K-means; gradient boosting; post-processing; mis-classification; machine learning

Paper 93: Optimizing Grape Leaf Disease Identification Through Transfer Learning and Hyperparameter Tuning

Abstract: Grapes are a globally cultivated fruit with significant economic and nutritional value, but they are susceptible to diseases that can harm crop quality and yield. Identifying grape leaf diseases accurately and promptly is vital for effective disease management and sustainable viticulture. To address this challenge, we employ a transfer learning approach, utilizing well-established pre-trained models such as ResNet50V2, ResNet152V2, MobileNetV2, Xception, and In-ceptionV3, renowned for their exceptional performance across various tasks. Our primary objective is to identify the most suitable network architecture for the classification of grape leaf diseases. This is achieved through a rigorous evaluation process that considers key metrics such as accuracy, F1 score, precision, recall, and loss. By systematically assessing these models, we aim to select the one that demonstrates the best performance on our dataset. Following model selection, we proceed to the crucial phase of fine-tuning the model’s hyperparameters. This fine-tuning process is essential to enhance the model’s predictive capabilities and overall effectiveness in disease identification. To accomplish this, we conduct an extensive hyperparameter search using the Hyperband strategy. Hyperparameters play a pivotal role in shaping the behavior and performance of deep learning models, and by systematically exploring a wide range of hyperparameter combinations, our goal is to identify the most optimal configuration that maximizes the model’s performance on the given dataset. Additionally, the study’s results were compared with those of numerous relevant studies.

Author 1: Hoang-Tu Vo

Author 2: Kheo Chau Mui

Author 3: Nhon Nguyen Thien

Author 4: Phuc Pham Tien

Author 5: Huan Lam Le

Keywords: Grape disease recognition; disease identification; transfer learning; hyperparameter optimization; hyperband strat-egy; fine-tuning; deep learning

Paper 94: An Internet of Things-based Predictive Maintenance Architecture for Intensive Care Unit Ventilators

Abstract: Intensive care units commonly utilize mechanical ventilators to treat patients with different medical conditions, which are crucial for patient care and survival. ICU ventilators have evolved through four distinct generations, each displaying unique features. Despite progress made since the 1940s, contemporary designs are insufficient to meet the increasing needs of patients and hospitals. Malfunctions in mechanical ventilators pose significant dangers to patients, highlighting the importance of focusing on their safety, security, precision, and dependability. Our study aims to address the significant issue at hand. Furthermore, the IoT industry has garnered significant attention because of rapid progress in smart devices, sensors, and actuators. The healthcare industry has seen a notable increase in health data as a result of the growing utilization of IoT and cloud computing technologies. To enhance growth, new models and distributed data analytics strategies must be developed to fully utilize the value of the vast datasets generated, including the incorporation of embedded machine learning. The study focuses on conducting Pareto and Failure Modes and Effects Analysis (FMECA) on ventilators in a specific hospital’s ICU, specifically those manufactured by the same company and unit. The analysis aims to identify the most critical and failure-prone component. Subsequently, we propose an IoT-focused framework for a predictive maintenance system implemented at the component level. The architecture comprises a monitoring framework and a data analytics module to predict potential system failures in advance, enhancing overall reliability.

Author 1: Oumaima Manchadi

Author 2: Fatima-Ezzahraa BEN-BOUAZZA

Author 3: Zineb El Otmani Dehbi

Author 4: Aymane Edder

Author 5: Idriss Tafala

Author 6: Mehdi Et-Taoussi

Author 7: Bassma Jioudi

Keywords: Internet of things; predictive maintenance; embed-ded Machine learning; data analytics; failure modes; mechanical ventilator

Paper 95: Predicting Aircraft Engine Failures using Artificial Intelligence

Abstract: Nowadays, the aviation sector continues to develop especially with the emergence of new technologies, and solutions. Hence, there is an increasing demand for enhanced safety and operational efficiency in the aviation industry. As to guarantee this safety, the aircraft’s engines must be monitored, controlled and maintained, however in an efficient way. Thus, the research community is working continuously in order to provide solutions that are efficient and cost effective. Artificial intelligence and more specifically machine learning models have been employed in this sense. Here comes the proposition of this article. It presents solutions implementing predictive maintenance using machine learning models. They help in predicting aircraft’s failures. This is in order to avoid operations of unscheduled maintenance and disruptions of services.

Author 1: Asmae BENTALEB

Author 2: Kaoutar TOUMLAL

Author 3: Jaafar ABOUCHABAKA

Keywords: Aircraft engine failures; machine learning; predic-tive maintenance; C-MAPSS; aviation safety

Paper 96: A Hybrid Model for Ischemic Stroke Brain Segmentation from MRI Images using CBAM and ResNet50-Unet

Abstract: Ischemic stroke is the most prevalent type of stroke and a leading cause of mortality and long-term impairment globally. Timely identification, precise localization, and early detection of ischemic stroke lesions brain are critical in healthcare. Various modalities are employed for detection, and magnetic resonance imaging stands out as the most effective. Different magnetic resonance imaging techniques have been proposed for the detection of ischemic stroke lesion tumors, allowing for image uploading and visualization. Automated segmentation of ischemic stroke lesions from magnetic resonance imaging images has an important role in the analysis, prognostic, diagnosis, and clinical treatment planning of some neurological diseases. Recently, computer-aided diagnosis systems based on deep learning techniques have demonstrated significant promise in medical image analysis, particularly in multi-modality medical image segmentation. Automated segmentation is a difficult task due to the enormous quantity of data provided by magnetic resonance imaging and the variation in the location and size of the lesion. In this study, we develop an automated computer-aided diagnosis system for the automatic segmentation of ischemic stroke lesions from magnetic resonance imaging images using a Convolution Block Attention Module (CBAM) and hybrid UNet-ResNet50 model. The UNet model is integrated into the architecture, and the ResNet50 backbone is pre-trained to enhance feature extraction. CBAM block is a model applied in this approach to extract the most effective feature maps. The proposed approach is evaluated on the public Ischemic Stroke Lesion Segmentation Challenge 2015 dataset, arranged into weighted-T1(T1), weighted-T2(T2), FLAIR, and DWI sequences. Experimental results demonstrate the efficacy of our approach, achieving an impressive accuracy value of 99.56%, a precision value of 97.12%, and a DC of 79.6%. Notably, our approach outperforms other state-of-the-art methods, particularly in terms of accuracy values, highlighting its potential as a robust tool for automated ischemic stroke lesion segmentation in magnetic resonance imaging.

Author 1: Fathia ABOUDI

Author 2: Cyrine DRISSI

Author 3: Tarek KRAIEM

Keywords: Medical image segmentation; ischemic stroke disease; UNet; ResNet50; convolution block attention module; magnetic resonance imaging; transfer learning

Paper 97: Optimizing Bandwidth Reservation Decision Time in Vehicular Networks using Batched LSTM

Abstract: Time-sensitive and safety-critical networked vehicular applications, such as autonomous driving, require deterministic guaranteed resources. This is achieved through advanced individual bandwidth reservations. The efficient timing of a vehicle decision to place a cost-efficient reservation request is crucial, as vehicles typically lack sufficient information about future bandwidth resource availability and costs. Predicting bandwidth costs often using time-series machine learning models like Long Short-Term Memory (LSTM). However, standard LSTM models typically require longer durations of multiple input data sets to achieve high accuracy. In certain scenarios, quick decisions must be made, even if the vehicle means sacrificing some accuracy. We propose a batched LSTM model to assist vehicles in placing bandwidth reservation requests within a limited data for an upcoming driving path. The model divides data during training to enhance computational efficiency and model performance. We validated our model using historical Amazon price data, providing a real-world scenario for experiment. The results demonstrate that the batched LSTM model not only achieves higher accuracy within a short input data duration but also significantly reduces bandwidth costs by up to 27% compared to traditional time-series machine learning models.

Author 1: Abdullah Al-khatib

Author 2: Klaus Moessner

Author 3: Holger Timinger

Keywords: Networked vehicular application; time-sensitive net-working; network reservation; batched LSTM

Paper 98: An Algorithm Based on Priority Rules for Solving a Multi-drone Routing Problem in Hazardous Waste Collection

Abstract: This research investigates the problem of assigning pre-scheduled trips to multiple drones to collect hazardous waste from different sites in the minimum time. Each drone is subject to essential restrictions: maximum flying capacity and recharge operation. The goal is to assign the trips to the drones so that the waste is collected in the minimum time. This is done if the total flying time is equally distributed among the drones. An algorithm was developed to solve the problem. The algorithm is based on two main ideas: sort the trips according to a given priority rule and assign the current trip to the first available drone. Three different priority rules have been tested: Shortest Flying Time, Longest Flying Time, and Median Flying Time. Two recharging conditions are maintained: recharging needed time and recharging full duration. By applying each priority rule and each recharging condition, we generate a six versions of the algorithm. The six versions of the proposed algorithm were implemented in Java programming language.The results were analyzed and compared proving that the Longest Flying Time priority rule surpasses the other two rules. Moreover, recharging a drone just enough for taking the next trip proved to be better than fully recharging it.

Author 1: Youssef Harrath

Author 2: Jihene Kaabi

Author 3: Eman Alaradi

Author 4: Manar Alnoaimi

Author 5: Noor Alawadhi

Keywords: Drones; trip assignment; priority rules; flying capacity; load balance

Paper 99: The Management System of IoT Informatization Training Room Based on Improved YOLOV4 Detection and Recognition Algorithm

Abstract: In response to the problems of low recognition rate and long system operation time in equipment detection management in the existing IoT information training room management system. A research has proposed an IoT information training room equipment detection management system on the ground of an improved YOLOV4 detection and recognition algorithm to solve the above problems. Firstly, it used the YOLOV algorithm to detect and identify equipment in the IoT information training room. Then, it used clustering methods to improve the YOLOV algorithm, thereby enhancing the detection accuracy and robustness of the algorithm, and thereby enhancing the performance of the equipment management system in the equipment management process of the training room. Finally, performance validation of the training room management system was conducted using datasets and simulation experiments. The results showed that the loss value of the training room equipment management system constructed using the improved YOLOv4 algorithm during the training process was 0.16. The accuracy and recall rates of device recognition were 95.71% and 92.83%, respectively. And the detection false alarm rate during the device detection and recognition process was only 2.15%, with a mAP value of 91.66%, and the detection and recognition indicators are higher than those of the comparison method. This indicates that the training room equipment management system constructed in the study has good adaptability in equipment detection and recognition in IoT information training rooms. The research aims to provide effective technical support for the management system of IoT training room equipment.

Author 1: Huiling Hu

Keywords: YOLOV4 algorithm; Internet of Things informatization; training room; management system; detection and recognition

Paper 100: Diagnosing Autism Spectrum Disorder in Pediatric Patients via Gait Analysis using ANN and SVM with Electromyography Signals

Abstract: Autism Spectrum Disorder (ASD) is a permanent neurological maturation condition that impacts communication, social interaction, and behavior. It is also associated with atypical walking patterns. This study aims to create an automated classification model to distinguish ASD children during walking based on the muscles Electromyography (EMG) signals. The study involved 35 children diagnosed with ASD and an equal number of typically developing (TD) children, all aged between 6 and 13 years. The Trigno Wireless EMG System was used to collect EMG signals from specific muscles in the lower limb (Biceps Femoris - BF, Rectus Femoris - RF, Tibialis Anterior - TA, Gastrocnemius - GAS) and the arm (Biceps Brachii - BB, Triceps Brachii - TB) on the left side. To identify the most significant features influencing walking in ASD children, a statistical analysis using the Mann-Whitney Test was conducted. The dataset contained 42 features derived from the analysis of six muscles across seven distinct walking phases throughout a single gait cycle. Following this, the Mann-Whitney Test was utilized for feature selection, uncovering five significantly distinctive features within the EMG signals between children with ASD and those who were typically developing. The most notable EMG features were subsequently employed in constructing classification models, namely an Artificial Neural Network (ANN) and a Support Vector Machine (SVM), aimed at distinguishing between children with ASD and those who were typically developing. The results indicated that the SVM classifier outperformed the ANN classifier, achieving an accuracy rate of 75%. This discovery shows potential for employing EMG signal analysis and classification model algorithms in diagnosing autism, thereby advancing precision health.

Author 1: Rozita Jailani

Author 2: Nur Khalidah Zakaria

Author 3: M. N. Mohd Nor

Author 4: Heru Supriyono

Keywords: Autism Spectrum Disorder; Electromyography signals; Artificial Neural Network; Support Vector Machine; precision health

Paper 101: Improving Load Balance in Fog Nodes by Reinforcement Learning Algorithm

Abstract: Fog computing is a distributed computing concept that brings cloud services out to the network's edge. Real-time user queries and data streams are processed by cloud nodes. Tasks should be evenly divided among fog nodes in order to maximize speed and efficiency, optimize resource efficiency, and reaction time. Real-time user requests and data flow processing are done by cloud nodes. Nodes in a network must share responsibilities in a balanced manner in order to maximize speed and efficiency, resource efficiency, and reaction time, hence in this article, a novel approach is presented. When it comes to fog computing, load balancing essential suggested to be improved. According to the suggested algorithm, a task submitted to the fog node via a mobile device would be processed by the fog node using reinforcement learning before being passed on to another fog node. Neighbor or let the cloud handle it. According to the simulation findings, the suggested algorithm has achieved a reduced execution time than other compared approaches by properly allocating the work among the nodes. Consequently, the suggested technique has reduced the chance of incorrect job assignment by 24.02% and the response time to the user by 31.60% when compared to similar methods.

Author 1: Hongwei DING

Author 2: Ying ZHANG

Keywords: Fog computing; resource allocation; reinforcement learning; delay; load balancing; fog nodes

Paper 102: Construction of an Art Education Teaching System Assisted by Artificial Intelligence

Abstract: With the continuous progress of art education and artificial intelligence technology, traditional music teaching models are facing transformation. This article aims to construct an art education and teaching system based on artificial intelligence, especially for teaching music sound recognition. Through in-depth research, we have designed a music sound recognition system that uses Mel frequency cepstral coefficient (MFCC) for feature parameter extraction, and combines BP neural network algorithm to construct a music sound learning model. The main purpose is to improve the efficiency and accuracy of music teaching through artificial intelligence technology. The main challenge we face in this process is how to effectively extract the features of music sounds and accurately identify different tones through algorithms. By using the MFCC algorithm, we have successfully solved this problem as it can effectively describe the time-frequency characteristics of music sound. Our proposed music sound learning model is based on a BP neural network, which trains the network to learn the mapping relationship between music sound and pitch. The experiment used piano sound as an example to verify the accuracy and reliability of the system. The simulation experiments conducted in MATLAB environment show that our system can accurately recognize and extract the main frequency of music, and has higher performance compared to traditional methods.

Author 1: Xianyu Wang

Author 2: Xiaoguang Sun

Keywords: Feature extraction BP neural network; Tone recognition; smart art teaching; MEL frequency cepstral coefficient; MFCC algorithm; time frequency characteristics

Paper 103: Design of Big Data Task Scheduling Optimization Algorithm Based on Improved Deep Q-Network

Abstract: Big data analysis can provide valuable insights not easily obtained from traditional data scales. However, addressing scheduling issues in big data can be challenging due to the vast amount and diverse nature of the data. To overcome this, a scheduling model based on Markov decision process is proposed. The deep Q-network algorithm is used for directed acyclic graph task scheduling. To improve this model further, the gradient strategy algorithm is introduced. From the results, when the dataset size was about 500, the hybrid algorithm achieved a recall rate of 0.96, outperforming the gradient strategy algorithm (0.83), deep Q-network algorithm (0.79), and estimated earliest completion time algorithm (0.63). Although the estimated earliest completion time algorithm had longer training times under different dataset sizes, the hybrid algorithm's training time was slightly longer than the gradient strategy algorithm and slightly shorter than the deep Q-network algorithm. Overall, the proposed algorithm exhibits superior performance and significant value in solving engineering problems.

Author 1: Fu Chen

Author 2: Chunyi Wu

Keywords: Big data; Task scheduling; Policy gradient; Deep Q-network