		FeedBack \| Contact Us \| Links \| Site Map

Welcome to Journal of Theoretical and Applied Information Technology

Submit Paper / Call for Papers

Journal receives papers in continuous flow and we will consider articles from a wide range of Information Technology disciplines encompassing the most basic research to the most innovative technologies. Please submit your papers electronically to our submission system at http://jatit.org/submit_paper.php in an MSWord, Pdf or compatible format so that they may be evaluated for publication in the upcoming issue. This journal uses a blinded review process; please remember to include all your personal identifiable information in the manuscript before submitting it for review, we will edit the necessary information at our side. Submissions to JATIT should be full research / review papers (properly indicated below main title).

Journal of Theoretical and Applied Information Technology
November 2024 | Vol. 102 No.21

Title:	DETECTION AND CLASSIFICATION OF DIABETIC RETINOPATHY USING YOLO-V8 DEEP LEARNING METHODOLOGY
Author:	GEETHA DEVI APPARI, T. V. HYMA LAKSHMI, K. VIDYA SAGAR, MORSA CHAITANYA, SRAVANTHI KANTAMANENI, VEERA VASANTHA RAO BATTULA, SURYA PRASADA RAO BORRA
Abstract:	Diabetic retinopathy (DR) is increasing rapidly. Early detection of DR using computer aided methodology is imperative. This paper aimed to detect various stages of DR meticulously. 60 percent of datasets are considered as trained data sets. 20 percent of datasets are considered as validation datasets. ‘20’ percent of the datasets are considered for testing. YOLOv8 is considered to extract the features and to detect the abnormalities of the fundus images. The edges information of the image is safeguarded with zero padding. Object detection, regression and classification is done with YOLOv8 methodology. The Convolutions apply filters to highlight specific patterns of the image. More spatial information is achieved with a minimum of 29 strides. But to increase the operational speed and to reduce the complexity in computation, the stride value is slightly increased to 31. The performance of the YOLOv8 is evaluated using performance metrics like Train loss function curve, precision recall curve, confusion matrix evaluation and F1 confidence curves. The training loss and validation loss are stabilized with increasing epochs to 25. The YOLOv8 classifier produced micro average precision (mAP) is 0.5 for all sets of results. This proposed methodology is significantly good for detecting diabetic retinopathy. The yolo 8 classifiers produced 0.5 mAP for all sets of results. The mAP value achieved for all classes is 0.5. For no diabetic retinopathy the F1 score value is 0.59 and for all classes F1 Score is 0.31 at 0.113. F1 Score is significantly good for CNN and YOLOv8 classification algorithms. i.e. 0.39 and 0.31. The predicted percentage for no DR is 54%. The predicted percentage for Mild DR is 38%. The prediction rate of proliferate DR is 71%. Based on the performance metrics this deep machine learning algorithm is deliberate as a stand-alone tool to detect the DR images.
Keywords:	Diabetic Macula Edema, OCT Images, Transfer Learning Models, ResNet-50, Diabetic Retinopathy, Medical Image Processing.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	INTEGRATED GLCM TEXTURE FEATURES AND CNN FOR AUTOMATED COTTON DISEASE IDENTIFICATION
Author:	POOJA MEENA, PRIYANKA GUPTA, DR.SOHIT AGARWAL
Abstract:	Background: Cotton diseases pose a significant threat to crop yield and quality, making timely and accurate identification crucial for effective management.Objectives: To introduce a novel methodology for the automated detection and classification of diseases affecting cotton plants by leveraging Gray-Level Co-occurrence Matrix (GLCM) feature extraction coupled with Convolutional Neural Network (CNN) classification.Method: The proposed approach involves extracting texture features from cotton plant images using GLCM, which captures crucial spatial relationships among pixel intensities. These features, encompassing contrast, correlation, energy, and homogeneity, serve as discriminative attributes for disease identification. Subsequently, a CNN-based classification model categorizes the extracted features into distinct disease classes, including cotton leaf curl virus (CLCV), bacterial blight, and healthy foliage.Findings:Experimental results demonstrate the efficacy of the proposed methodology, achieving high accuracy (87%), precision (0.88), recall (0.85), and F1-score (0.87) in cotton disease detection.Novelty: The integration of GLCM feature extraction with CNN classification offers a promising solution for the automated and precise diagnosis of cotton diseases, facilitating early intervention and sustainable management practices in agriculture.
Keywords:	Cotton Diseases, Gray-Level Co-occurrence Matrix, Convolutional Neural Network, Automated Detection, Disease Classification, Texture Features, Agricultural Management.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	ENHANCING USER SAFETY: AI-DRIVEN POPUP ALERTS FOR SUSPICIOUS CONTENT USING DEEP MULTI-MODAL ANOMALY FUSION NETWORK
Author:	B SHANTHINI, Dr. N SUBALAKSHMI
Abstract:	This research introduces a novel approach to enhancing user safety in online environments through AI-driven popup alerts for detecting suspicious content. By using a Deep Multi-Modal Anomaly Fusion Network, we integrate features extracted from text, emojis, images, and videos to construct a comprehensive representation of product reviews across diverse modalities. In particular, we employ the innovative AdaptiMatrixFactorizer for feature extraction, which dynamically adjusts its factorization process to capture evolving data patterns. The proposed fusion network architecture seamlessly integrates features from different modalities, utilizing concatenation, element-wise addition, and attention mechanisms to facilitate effective multi-modal anomaly detection using Deep Multi-Modal Anomaly Fusion Network (DMMAFN). Furthermore, our approach introduces a dynamic threshold adjustment mechanism within the fusion network to adaptively regulate anomaly detection sensitivity based on real-time changes in data patterns and distributions. This adaptive thresholding strategy considers three critical parameters: sentiment analysis, repetitive reviews, and spatio-temporal analysis. Upon detection of suspicious content exceeding the dynamically adjusted threshold, a popup alert is generated to notify users, fostering a safer online environment. Extensive experimentation and evaluation on anomaly-labeled datasets demonstrate the efficacy and reliability of our approach in accurately detecting and alerting users to potential risks in product reviews across various online platforms. This research contributes to advancing user safety in online environments by providing a proactive and intelligent solution for identifying and addressing suspicious content, while also introducing the novel AdaptiMatrixFactorizer for dynamic feature extraction.
Keywords:	AI-driven Popup Alerts, Suspicious Content Detection, Deep Multi-Modal Anomaly Fusion Network, AdaptiMatrixFactorizer, Sentiment Analysis
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	EMPATHIZING PARKINSON'S DISEASE USING SUPERVISED LEARNING METHODS
Author:	SUMALATHA MADUGULA, ARAVIND GUNDA, YANDRA SRINIVAS, N.C.KOTAIAH, G.NAGA RAJU, SIVA GANGA BADIPATI, BORRA BHAVITHA
Abstract:	Parkinson's disease (PD) is a common neurological illness that emphasizes how important it is to identify symptoms early for effective treatment. The primary cause of Parkinson's disease is the gradual degeneration of dopamine-producing neurons in the brain. This study looks into the application of machine learning models for PD recognition, including XGBoost, CatBoost, Decision Tree, Random Forest, K-Nearest Neighbours, and LightGBM. Parkinson's disease is very common, which emphasizes how important it is to diagnose the condition early and begin treating it quickly. The robustness of the model is enhanced by Logistic Regression, KNN, Decision Tree, Random Forest, CatBoost, and XGBoost; nonetheless, LightGBM outperforms the other models with an accuracy rate of 96.52%. This work not only provides a valid method for diagnosing Parkinson's disease early on, but it also sheds light on the potential for combining several machine learning models to improve diagnostic performance, which will ultimately lead to improvements in neurology and personalized medicine.
Keywords:	Machine Learning Models, Diagnostic, Health care, KNN
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	SURVEYING THE BLACK BOX: AN OVERVIEW OF CONTEXT-AWARE RECOMMENDER SYSTEMS
Author:	KAOUTAR ERRAKHA, AMINA SAMIH, ABDERRAHIM MARZOUK, AYOUB KRARI
Abstract:	Researchers and practitioners across various fields such as data mining, marketing, management, mobile computing, and personalized e-commerce have increasingly recognized the importance of contextual information in refining user experiences. Traditional recommender systems, however, tend to focus basicly on recommending the most relevant items to users, often overlooking essential contextual factors like time, location, or social context (e.g., dining out with friends or watching a movie). Addressing this limitation, recent advancements in context-aware recommender systems (CARSs) leverage contextual data to improve recommendation quality, gaining considerable traction in the research community. In this paper, we present an overview of CARS, tracing its evolution and the diverse approaches that have emerged over recent years. We provide a critical review of recent context-based recommendation methods to highlight prevailing trends and limitations. From this analysis, we then identify supplementary features that could enhance context-aware recommender systems. Finally, we explore promising research directions that could overcome current limitations, encouraging experts to collaborate in advancing context-aware recommendation technologies further.
Keywords:	Recommender Systems, Context-aware recommendation systems, CARSs, Context, Contextual Information
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	EMERGING TREND OF AMBIDEXTROUS AI-DRIVEN TECH VENTURES
Author:	SARA AZIZ, NOORLIZAWATI ABD RAHIM
Abstract:	Artificial Intelligence (AI) as a potent catalyst revolutionized legacy business models and led to surge in AI-driven tech ventures. However, there is a gap in understanding of the foundational factors for their emergence, opportunities and challenges they face, and the future trends likely to shape their growth. This paper aims to bridge that gap by employing multiple theoretical underpinnings including Technology Acceptance Model, Diffusion of Innovations Theory, Ambidexterity Innovation, and the Resource-based View to investigate AI integration within these ventures. Using Systematic Literature Review (SLR) approach, 41 studies from WOS and SCOPUS databases covering the period 2020-23 were analyzed and identified four key drivers including technological advancements, market demand, investment and innovation promotion. We conceptualize 'Ambidextrous AI-driven Tech Ventures' as businesses that integrate AI technologies to simultaneously explore new market opportunities and exploit existing resources. However, these ventures face several challenges including skill demands, ethical dilemmas, cyber security risks, and regulatory hurdles. Findings emphasize the need for strategic stakeholder support and targeted investment in developing these ventures' dual capabilities required for sustainable growth. This study offers actionable insights for entrepreneurs, investors, policymakers, and academics stress the urgency of ongoing research and education to navigate and capitalize on sustainable development within AI-driven businesses effectively.
Keywords:	Ambidexterity, Artificial Intelligence, Tech Ventures, Digital Innovation, Strategic Entrepreneurship, Exploration, Exploitation
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	NAVIGATING CYBERSECURITY: A COMPREHENSIVE ANALYSIS OF MACHINE LEARNING IN CYBER ATTACK DETECTION
Author:	S. DEEPA RAJAN, A. MANIKANDAN
Abstract:	In the ever-evolving landscape of cyber threats, the integration of machine learning (ML) techniques has emerged as a powerful tool for detecting and mitigating attacks across various sectors, such as the Internet of Things (IoT) and Wireless Sensor Networks (WSN). This analysis paper examines several ML algorithms, such as Random Forest (RF), Ridge Classifier, and Gaussian Naive Bayes, and their efficacy in enhancing cyber-attack detection accuracy and efficiency. Emphasising the significance of preprocessing data and feature extraction, our paper highlights the exceptional performance of hybrid models and the transformative role of Multi-Agent Reinforcement Learning (MARL) in addressing the challenges posed by class imbalance and rapidly evolving threats. This paper underscores the critical need for continuous innovation in cybersecurity measures by showcasing the promising results achieved through these techniques. Ultimately, our findings reveal that while various ML approaches have successfully detected cyber-attacks, the implementation of MARL represents a significant advancement in developing robust and adaptive intrusion detection systems.
Keywords:	Cyber-Attack, Machine-Learning, Multi-Agent Reinforcement Learning, Cyber-Security, Intrusion Detection System.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	EVALUATING RNNS AND TRANSFORMERS FOR CODE-RELATED TASKS INCLUDING BUG DETECTION, CODE COMPLETION, AND SUMMARIZATION
Author:	RAGHUPATHY DURGA PRASAD, Dr. MUKTEVI SRIVENKATESH
Abstract:	Software engineering tasks such as bug detection, code completion, and code summarization are critical for improving the efficiency and reliability of software development. With the increasing complexity of modern codebases, there is a need for robust deep learning models capable of understanding and predicting patterns in source code. This study addresses the challenge of selecting the most suitable model for these tasks by conducting a comprehensive evaluation of two prominent architectures: Recurrent Neural Networks (RNNs) and Transformers. RNNs are effective at capturing short-term dependencies but struggle with long sequences, while Transformers excel at modelling long-range dependencies through self-attention mechanisms. This research evaluates the performance of these models across three key tasks—bug detection, code completion, and summarization—using multiple datasets. The results demonstrate that Transformers consistently outperform RNNs in terms of accuracy and Bilingual Evaluation Understudy (BLEU) scores, particularly for tasks involving long code sequences, while RNNs are more computationally efficient in memory-constrained environments. The study contributes to the field by providing practical insights for developers and researchers on how to leverage these models based on task requirements and available computational resources. These findings highlight the potential of Transformers to enhance the accuracy of software engineering tools, while also presenting a trade-off in terms of resource consumption, making this study valuable for future model selection and optimization efforts.
Keywords:	Source Code Analysis, Bug Detection, Code Completion, Code Summarization, Recurrent Neural Networks (RNNs), Transformer Model
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	TOWARDS EFFICIENT PATENT CLASSIFICATION: KOLMOGOROV ARNOLD NETWORKS AS AN ALTERNATIVE TO MLP
Author:	MINJONG CHEON, CHANGBAE MUN
Abstract:	In todays continuously shifting innovation and technological growth environment, effective intellectual property (IP) management and organization have become critical, resulting in more significant patent classification. Moreover, recent advances in natural language processing (NLP) technology have resulted in enhanced patent categorization. However, incorporating multilayer perceptron (MLP) layers in NLP algorithms frequently results in higher memory needs, particularly as network size rises. We suggest using the Kolmogorov Arnold Network (KAN) instead of MLP layers to solve this issue. In this work, we used a dataset from the European Patent Office (EPO) to categorize patents into three groups. We experimented with several KAN setups and discovered that decreasing hidden dimension sizes considerably reduced the number of parameters while keeping good accuracy. The [32, 16, 8] configuration achieved an accuracy of 74.84%, which rose to 75.12% after adjusting crucial hyperparameters such as spline_order and grid_size. Compared to other machine learning models such as MLP (75.83%), Random Forest, and XGBoost, KAN consistently surpassed them in accuracy and efficiency. Our findings broaden the use of KAN to patent classification and offer new avenues for its usage in other text-based classification tasks. KAN's proven efficiency and performance make it a promising alternative to existing machine learning models in this area, emphasizing its potential for further application in patent-related activities.
Keywords:	Patent classification, NLP, Kolmogorov Arnold Network (KAN), MLP, Deep Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	PALM OIL PRODUCTION PREDICTION USING LONG SHORT-TERM MEMORY METHOD
Author:	HANDRIZAL, DEWI SARTIKA GINTING, HERRIYANCE, SYABRINA RAMADHANI KAMAL
Abstract:	Palm oil production is one of the important indicators in the agribusiness sector. Accurate prediction of palm oil production helps better planning and decision-making. This study's problem is building a palm oil production prediction model using Long Short-Term Memory (LSTM). The LSTM model consists of four layers with 50 memory units and 20% dropout. The model is compiled using the RMSprop optimizer and the RMSE loss function. The evaluation results show that the LSTM model provides palm oil production predictions with RMSE of 0.1360 for total FFB received, 0.1279 for total FFB processed, and 0.1279 for total CPO which shows good potential. The contribution of this research can provide knowledge about predicting palm oil production so that companies can plan more efficient production.
Keywords:	Long Short-Term Memory (LSTM), Palm Oil Production Prediction, MinMaxScaler, Root Mean Square Error (RMsE), Deep Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	BEACON SIGNAL AND SLEEP AND AWAKE STRATEGY IN MANET FOR POWER ENHANCEMENT
Author:	DR.S.HEMALATHA, KHADRI SYED FAIZZ AHMAD, VENKATA RAO TAVANAM, R.V.V. KRISHNA, U. SATHYA, DR.N.MUTHUVAIRAVAN PILLAI
Abstract:	In Mobile Adhoc Networks (MANETs), efficient battery management is crucial for maintaining device longevity and ensuring stable network communication. A device's battery depletion during active communication can disrupt the entire network and necessitate packet retransmission, causing inefficiencies. Although various strategies, such as optimal routing protocols, optimization techniques, and duplicate packet elimination, have been explored to manage battery power, these approaches still require specialized power-saving mechanisms. This study introduces a novel strategy using beacon signals combined with sleep and awake cycles for network nodes to optimize power consumption and enhance communication reliability. A cluster-based approach is adopted, where a designated Cluster Head (CH) is responsible for managing beacon signals and coordinating sleep and awake schedules across the nodes. This method, termed CHSAB-AODV (Cluster Head Sleep-Awake Beaconing with AODV), was evaluated using Network Simulator 3 (NS3) and compared against traditional AODV implementations. The simulation results indicate that CHSAB-AODV outperforms standard AODV protocols, reducing power consumption by 25% to 50% while maintaining communication efficiency. This significant improvement highlights the effectiveness of beacon-based sleep and awake strategies in addressing power management challenges in MANETs.
Keywords:	MANET , Beacon Signal, Sleep And Awake , Power Management
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	IMPROVING INTERACTIVE MEDIA EDUCATION: INVESTIGATING THE CAPABILITIES OF EDUCATIONAL PODCASTING PLATFORMS FOR ENHANCED LEARNING
Author:	SARNI SUHAILA RAHIM, SHAHRIL PARUMO, KHAIRUL YASSER JUSNI, SURIATI KHARTINI JALI
Abstract:	This study examines the transformative potential of interactive educational podcasting platforms in higher education. These platforms employ digital technology to enhance learning experiences and address educational shortcomings. The study evaluates a specialized interactive educational podcasting platform designed to support higher education students by offering a variety of subjects and modules tailored for distinct disciplines in interactive media courses. An experiment involving 19 participants comprising subject matter experts, multimedia experts, and students was conducted to assess the platform’s effectiveness as a learning tool. The results indicate a positive impact on student engagement and learning outcomes. Interactive educational podcasting platforms provide personalized, accessible, and engaging learning experiences, presenting a significant opportunity to revolutionize higher education. Future research and collaboration will be crucial for optimizing these platforms and maximizing their potential to transform educational practices.
Keywords:	Podcast, Web Development, Interactive Media, Teaching and Learning, Higher Education
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	LEVERAGING ARTIFICIAL INTELLIGENCE FOR SMES' SUSTAINABLE COMPETITIVE ADVANTAGE: THE MODERATING ROLE OF MANAGERS DIGITAL LITERACY
Author:	QADRI KAMAL ALZAGHAL, OMAR HASAN SALAH, MOHANNAD MOUFEED AYYASH
Abstract:	In the modern digital era, SMEs have become one of the most important sectors within the economy, and they now face significant pressure to keep up with their competitors while incorporating sustainability into their operations, especially in Palestine. In this context, the role of digital literacy among managers in moderating SMEs' sustainable competitive advantage is paramount. This research investigates the possibility of adopting AI to enable SMEs in Palestine to achieve a sustainable competitive advantage. It explores how the AI-driven approach can optimize business processes, enhance decision-making, and fuel innovation under resource-constrained conditions. The study emphasizes the crucial role of digital literacy in ensuring the successful integration of AI technologies into the company's activities. This study will be critical in determining the drivers of AI adoption and the impact on Sustainable performance in Palestinian SMEs. A sample of 284 SMEs was drawn from the questionnaire using a simple random sampling technique. Data were analyzed using partial least squares-structural equation modeling to test the relationships among the exogenous, moderator, and endogenous variables. Empirical analysis shows that managers' digital literacy significantly enhances the effectiveness of AI adoption and is, therefore, instrumental in enabling SMEs to exploit AI for long-term sustainability and growth. These findings bring essential insights for policymakers and business leaders in developing targeted training programs to address the digital literacy gap and promote AI adoption in Palestinian SMEs.
Keywords:	Artificial Intelligence; SMEs; Sustainable Competitive Advantage; Managers Digital Literacy; Palestine.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	DEVELOPMENT OF INTERACTIVE FORMS OF COMMUNICATION THROUGH ARTIFICIAL INTELLIGENCE
Author:	OLEKSII SYTNYK, OLENA HROZNA, DANYLO FILONENKO, MARIANNA KONOPLYANNIKOVA, OLHA SERDIUK
Abstract:	Artificial intelligence (AI) has opened up new opportunities for implementing interactive media communication in the digital environment, demonstrating its potential. However, developing users’ conscious, responsible attitude to interaction with AI tools, critical thinking, and technological knowledge requires a complex and balanced pedagogical approach. The aim of this article is to investigate the potential of AI tools to increase the effectiveness of interactive forms of communication in the digital environment. The research employed the following empirical methods: experiment, questionnaire, as well as qualitative and quantitative analysis. The concept of developing interactive media communication skills was developed. An educational experiment was conducted to analyse the capabilities of AI tools in increasing the effectiveness of interactive forms of communication in the digital environment. The results of the survey of students taking the experimental course showed that technologies are most effective in the following processes: creation and editing of interactive media content (4.89 points), distribution of content in the digital environment (4.76 points), fact-checking (4.27 points), and data collection for case studies (4.03 points). The results of the survey of the participants of the educational experiment - a group of media content consumers showed that interactive communication in the digital environment left a positive impression on the respondents. The respondents rated their experience as creative (9.7 points), exciting (9.5 points), informative (9.3 points), interesting (8.9 points), inspiring (8.8 points), intriguing (8.5 points), stimulating (8.1 points). The article may be of interest to educators who are looking for optimal strategies for integrating AI tools into educational programmes for the development of interactive media communication skills in a digital environment.
Keywords:	Artificial Intelligence (AI) Tools, Media Industry, Media Content, Interactive Communication, Content Generation, Fact-checking, Content Distribution
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	DIGITAL EMPOWERMENT AND CYBERSECURITY: UNDERSTANDING PUBLIC AWARENESS OF DIGITAL INDIA INITIATIVES
Author:	RICHA SHARMA, ASTHA GOYAL, ROLI BANSAL, CHETAN YADAV3, LIPIKA, ESHAAN R JAMES
Abstract:	Digitalization refers to the process of adopting digital technologies to streamline and enhance everyday activities, thereby making life easier. This includes the use of digital tools to facilitate communication, access to information, transactions and service delivery, which improves the efficiency, convenience and overall user experience. This study primarily examines the daily usage of digital technologies and cyber awareness among individuals in general, across different age groups and other demographics in the Indian context. In this research, we examined the adoption of digital technology, by focusing on the usage of financial services, online payment portals, and involvement in government initiatives. A survey was conducted to collect data about the digital awareness and cyber-attack experiences of the participants. A thorough analysis of the response data shows that there is a stronger level of interaction with digital platforms and cyber awareness within urban adults in comparison to other age groups and areas of residence. It has also been observed that digitalization of basic services has made life easier for a large section of society, still there is a need for extensive promotion of government initiatives and online resources available in specialized domains like education, healthcare and business. Also, there is a scope for improvement regarding spreading digital awareness through educational initiatives and training programs and effective and timely redressal of complaints pertaining to cybercrimes.
Keywords:	Digitalization, Digital Literacy, Technology Access, Digital India, Cybersecurity, Cyber Awareness, Digital Adoption
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	AN ENHANCED EARLY DETECTION AND RISK PREDICTION OF BRAIN TUMORS USING MRI-CT SCANS WITH DEEP LEARNING TECHNIQUE
Author:	DINESH VALLURU, K. VARADA RAJKUMAR, PRAVEEN KUMAR K, BAJJURI USHARANI, Dr. E. SREEDEVI, J V RAMA KUMAR, PRADEEP JANGIR, Dr. SIVA KUMAR PATHURI
Abstract:	Big data analytics, like deep learning, is a flourishing technology in the medical field. This combination has the potential to influence how tumor illnesses are predicted, monitored, diagnosed, and treated. Brain tumors are regarded as the deadliest kind of tumors due to their fast development, shorter life duration, and diverse features. Misdiagnosis or insufficient health treatment might further diminish the likelihood of survival. The intricacy of brain tumors makes it impossible to distinguish them from normal tissues, making diagnosis challenging. Accurate diagnosis is critical for justifying therapy efficacy and patient survival over time. Despite extensive research into the process of identifying brain tumors, it remains a very challenging task due to the uneven distribution of lesions throughout several anatomical locations. The infrequent locations with unusual lesion distribution challenge categorization since these processes are present in ordinarily seeming tiny parts. Early and precise diagnosis of brain tumors is critical to delivering effective therapy and improving survival rates. Advances in medical imaging and deep learning techniques have integrated, making computerized brain tumor picture segmentation and classification increasingly viable. As a result, this study introduces a new deep learning model aimed at segmenting and classifying brain cancers from MRI-CT images. The Proposed model employs Res-Net50 for feature extraction (classification), followed by an ensemble model-based classifier and compared with the two base classifiers like Support Vector Machine, or SVM, and Decision Tree, to improve accuracy. The suggested classifier has a higher accuracy rating than the other two classifiers (95%).
Keywords:	Brain Tumor, ResNet50, Ensembled Classifier, Early Detection, Support Vector Machine, Decision Tree.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	INTELLIGENT SYSTEM FOR PRECISION TREATMENT MANAGEMENT FOR PATIENTS WITH BREAST CANCER METASTASIS
Author:	SARAH KHROUCH, LOUBNA CHERRAT, MOSTAFA EZZIYYANI
Abstract:	This study focuses on developing a novel decision-making system for oncologists to select the most suitable treatment strategy for breast cancer patients. The proposed system, called the Hybrid Model Process of Treatments (HMPT), is designed to assist oncologists by incorporating patient history, experiences, and responses. We analyzed clinical, digital pathology, and genomic data from patients treated with neoadjuvant chemotherapy to construct a strategic approach for metastatic breast cancer treatments. The HMPT model integrates two components: a Predictive Model (PM) using Neural Network (NN) and Logistic Regression (RL) to accurately forecast treatment outcomes, and a Data Augmentation Model (DAM) that generates new data. This newly generated data is evaluated against the Predictive Model (PM) to ensure alignment with established patterns. Results demonstrate that the model can be applied effectively across various breast cancer types, showing potential to expand clinical trial evaluations and test novel hypotheses for metastatic breast cancer patients. The HMPT model offers a revolutionary approach to reducing recurrence rates and enhancing the treatment experience, while also lowering the associated healthcare costs for patients.
Keywords:	Breast Cancer, Treatment management intelligent system, LR, NN, GA.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	JUMBLEDKEYS: TWO FACTOR USER AUTHENTICATION SCHEME USING PARTITIONED VIRTUAL KEYBOARD
Author:	S. RAJARAJAN, PLK. PRIYADARSINI
Abstract:	Passwords are the oldest and most widely used method of user authentication. While password based authentication is cost-effective, easier, faster and flexible, it is also susceptible to attacks. Particularly at the client device, passwords are most vulnerable when users enter their passwords. Before the password is encrypted and forwarded to the server, it can be captured by the attackers by exploiting the loopholes at the client systems that users use. In this paper, we introduce a secure virtual keyboard scheme designed to shield passwords from attackers during entry. The keyboard is partitioned into four groups and the keys in each group are randomized after each password character entry. Instead of directly clicking on the actual password characters, users will be clicking on the designated target keys as per a key-transfer scheme. The key-transfer scheme will be communicated to users through a SMS to their registered mobile number every time they attempt to login, effectively making the mobile phone a second factor of authentication. The jumbledKeys keyboard then generates a dynamic password based on the keys clicked by users. Only this dynamic password is stored at the client’s form and sent to the server. Attackers cannot trace the actual password without the knowledge of the key transfer scheme and the position of keys on the keyboard. This protects passwords against the shoulder-surfing, form grabbing, public wifi and man-in-the-middle attacks. Our user survey results proved that our scheme has not compromised on usability while elevating security
Keywords:	User authentication, Password attacks, Internet banking, Shoulder-surfing, Form grabbing, Keylogging, Virtual keyboard, Cyber security
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	COMBINATION OF GALE-SHAPLEY AND PESA-II ALGORITHM IN STUDENT-ACCOMMODATION MATCH
Author:	TRINH BAO NGOC, DANG NHAT QUANG, DANG TIEN DAT, LE ANH PHAN4, NGUYEN NGOC PHUONG KHANH, HA THI THANH THAO, HOANG PHUONG THAO, LE THI CHUNG
Abstract:	The inability of new students to secure suitable accommodation not only diminishes their quality of life and academic performance but also under- mines the reputation of the university. This paper tackles the problem by using Stable matching theory, a mathematical framework facilitating mutually beneficial relationships over time. By combining the Gale-Shapley algorithm and the PESA-II algorithm which is a multi-objective evolutionary optimization method, our approach systematically evaluates each student's requirements for accommodation, checking based on the recommendation list provided by the university for each student, striving to create a stable and fulfilling match. This guarantees a fair result that matches the preferences of both students and accommodations, with two main outcomes: students gain access to optimal accommodations, meeting all requirements, fostering an improved academic environment.
Keywords:	Student Selection, Boarding House Arrangement, Stable Matching Theory, Gale-Shapley Algorithm, PESA-II Algorithm.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	ADVANCED LIGHTWEIGHT ST-TCN FRAMEWORK USING UAV MULTI-SPECTRAL REMOTE SENSING FOR SURVEILLANCE AND CONTROL OF PINE NEMATODE DISEASE
Author:	Dr. J. DEEPA, Dr. V. GOKULA KRISHNAN, Dr. S. VENKATA LAKSHMI, Dr. D. ARUL KUMAR, Dr. V. VIJAYARAJA
Abstract:	The pine nematode is a highly infectious disease that devastates pine forests globally, necessitating prompt and precise diagnostic approaches. However, existing methods face challenges in accurately identifying and localizing nematode infections within individual trees, with limited studies addressing these gaps. This research contributes a novel artificial intelligence-based approach that leverages multi-spectral remote sensing imagery captured by unmanned aerial vehicles (UAVs) to diagnose pine nematode infections. By utilizing lightweight special task temporal convolutional network (ST-TCN) blocks and multiple bottleneck layers, our model focuses on critical features in the input sequence. This enables the classifier to differentiate between various diseases effectively. We further improve classification accuracy by fine-tuning model parameters through a Cat Swarm Updated Black Widow (CSUBW) optimization algorithm. The proposed method offers a rapid, accurate, and practical solution for monitoring and managing pine wood nematode disease, marking a significant advancement in the detection and control of this infection.
Keywords:	Unmanned Aerial Vehicle, Special Task Temporal Convolutional Network, Cat Swarm Updated Black Widow Model, Remote Sensing Images, Pine Nematode Disease.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	PROACTIVE SAFETYNET: A MOBILE-BASED COLLISION DETECTION AND ALERT SYSTEM FOR ENHANCED ROAD SAFETY
Author:	SOUSSI NIAIMI Badr-Eddine, BOUHORMA Mohammed, ZILI Hassan
Abstract:	Road accidents represent a significant global challenge, causing severe human casualties and economic losses. This research addresses the need for proactive safety measures by developing "Proactive SafetyNet," a comprehensive mobile-based solution to improve road safety. The system consists of two applications: one for drivers, which detects sudden jolts or shakes that may indicate an accident and, if the driver is unresponsive, sends automated alerts to nearby drivers; and a second app for authorities, which organizes incident alerts based on proximity and response time to ensure rapid intervention. By leveraging mobile technology and crowdsourced data, "Proactive SafetyNet" offers a participatory approach to road safety, aiming to reduce the impact and frequency of accidents through timely notifications and coordinated response.
Keywords:	Traffic Accidents, Proactive Safety Net, Mobile Applications, Shake Detection, Safety of Drivers, Response During Emergency, Crowdsourced Data, Prevention of Accident, Proximity-Based Warning, Road Safety in Collaboration, Economic Losses, Human Casualties, Mobile Technology, Rapid Response, Incident Reporting, User Engagement.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	HYBRID DEEP LEARNING AND SVM FOR BIOMARKER OF COLORECTAL CANCER TISSUE DECOMPOSITION
Author:	ITO WASITO1, HANDRI SANTOSO, DENNY SAPTONO F, EKAN FAOZI
Abstract:	Colorectal Cancer (CRC) is the third most common form of cancer and the second deadliest disease. The development of targeted therapy for cancer treatment increased demand for identification of molecular targets, such as driver mutations in cancer cells. However, some molecular tests, including next-generation sequencing, are not available to all cancer patients because of high cost and technical barriers. Standard biomarkers often pertain to costly and slow genetic tests. In recent development, Hematoxylin and eosin-stained biopsy slides are regularly available for colorectal cancer patients. Fortunately, rapid development has shown that objective biomarkers can be extracted from these images using Deep Learning (DL) approaches especially convolutional neural networks. This report proposes the hybrid method based on Deep Learning with CNN architecture as feature extractor and SVM as classifier to decompose nine classes of colorectal carcinoma slides images. This research has two main contributions, first, this research can provide insight to medical expert and computer scientist related to the current state of development deep learning based approaches for histopathological images classification especially in colorectal carcinoma (CRC). The second contribution of this research proposes alternative method in framework development of biomarker decomposition of CRC feature extraction, feature selection and classification of slides images. The results show that the proposed method has accuracy between 95%-99.5% in training data set and 94%-98.5 in the external data set. The biomarker of CRC classes LYM, STR and TUM successfully had been decomposed with high percentage of decomposition
Keywords:	Deep Learning, Convolutional Neural Networks, Support Vector Machine, Colorectal Carcinoma, Cancer Biomarker
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	DLMF: AN INTEGRATED ARCHITECTURE FOR HEALTHCARE DATA MANAGEMENT AND ANALYSIS USING DATA LAKE, DATA MESH, AND DATA FABRIC
Author:	LAMYA OUKHOUYA, ANASS EL HADDADI, BRAHIM ER-RAHA1, ASMA SBAI
Abstract:	The vast amount of data on healthcare, combined with the diversity of diseases, has led to a proliferation of work aimed at designing decision-making architectures capable of exploiting this information. These architectures are based on integrating heterogeneous data from different sources while ensuring that it is stored centrally. They also focus on data quality to guarantee the accuracy of analyses, and on reliable governance to ensure data compliance, security and traceability. These aspects are essential to enable optimal use of data for advanced analysis and informed decision-making in healthcare. Our objective in this article is to propose an architecture that ensures the 3 points: integration, storage and governance. The article proposes the DLMF architecture to ensure that these 3 points are adapted to good data analysis. This architecture uses the concept of a data lake for the consolidation and storage of data sources, a data mesh and a data fabric to ensure everything to do with data integration, quality and governance. The article also proposes a set of open-source technologies for its implementation. Finally, recommendations and future directions are suggested for a well-designed BI architecture capable of ensuring good data management, from data integration to analysis.
Keywords:	Data Lake, Data Mesh, Data Fabric, Healthcare, Decision Support Architecture
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	ANALYSIS OF THE SECURITY DEVELOPMENT OF BUSINESS ENTITIES IN THE CONDITIONS OF ARTIFICIAL INTELLECTUALIZATION OF THE GLOBAL SPACE
Author:	SVITLANA TULCHYNSKA, ANNA POHREBNIAK, OKSANA ZYBAREVA, МARIIA POZHYDAIEVA, OLEKSANDR SOLOSICH, ANDRII VAKULENKO
Abstract:	Evaluating the security development of business entities requires compliance with the criteria of complexity and systematicity in terms of taking into account the multi-faceted spectrum of internal functional and process relationships of the security configuration of the enterprise management, which requires the selection of a wide range of the indicators of parametric evaluation. The aim of this article is to justify a methodical approach to assessing the level of economic security of business structures in the conditions of artificial intellectualization of the global space using the method of taxonomic analysis as a tool for assessing the level of economic security. To use the method of taxonomic analysis in the framework of assessing the level of economic security of economic entities in the conditions of artificial intellectualization of the global space, in the article, the authors substantiated a system of analytical parameters describing the key components of ensuring economic security of economic entities in the context of artificial intellectualization processes of economic development. Differentiation of analytical parameters by functional characteristics of influence on the level of economic security, formation of the corresponding intellectual evaluation component was carried out. The proposed methodical approach was tested on the example of PJSC "Southern Mining and Processing Plant". The applied value of the proposed methodical approach lies in the possibility of substantiating the tools of strategic planning of the company's activities in the conditions of intensification of crisis phenomena and the availability of opportunities to overcome them by means of the intellectualized development.
Keywords:	Artificial Intellectualization, Economic Security, Enterprise, Industrial Security, Personnel Security, Financial Security, Intellectual Security
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	AUTOMATIC CLASSIFICATION OF ECG AND PCG SIGNALS USING CONVOLUTION NEURAL NETWORK FOR DETECTING CARDIOVASCULAR DISEASE
Author:	P JYOTHI, G. PRADEEPINI
Abstract:	Cardiovascular disease increasing deaths worldwide effecting young age people at their early stage. Heartbeat analysis of a person can be normal or abnormal heart sounds which can be detected only through trained physician. To reduce the dependency on trained physicians for heart sound detection, proposed system focuses on automatic classification of ECG(Electrocardiogram) and PCG (Phonocardiogram) signals after removal of unwanted signals and noise can be eliminated through filters. Convolution Neural Networks to eliminate manual extraction of the features of ECG and PCG signals. Classification of heart disease to detect heart rate by combining signals of ECG and PCG of electrical and mechanical activity to analyse signals effectively this paper proposes R peak detection using convolution method which is a mathematical way of combining two signals to form into new signal in digital signal processing is an efficient technique to detect heart rate through ECG and PCG signals for cardiac disease. The main moto of proposed work is to make use of both the ECG and PCG signals to detect R peaks from both the combined signals of ECG and PCG to detect heart rate abnormality. Thus, the proposed system helps the physicians for automatic classification to diagnose the heart rate in an efficient manner.
Keywords:	Cardiovascular, Electrocardiogram, Phonocardiogram, Convolution Neural Networks, Arrhythmia
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	ADVANCED CREDIT CARD FRAUD DETECTION: A NOVEL APPROACH INTEGRATING ADVERSARIAL-GUIDED OVERSAMPLING WITH MACHINE LEARNING
Author:	ABDEL RAHMAN AMR, WAEL HASSAN GOMAA*, FARID ALI MOUSA
Abstract:	Fraud detection in credit card transactions presents a significant challenge due to the highly imbalanced nature of the data, where fraudulent transactions make up only a small fraction of the total. In this paper, we introduce a novel approach to address this issue by integrating adversarial-guided oversampling with machine learning techniques. Our method enhances the detection of fraudulent transactions by focusing on the minority class, using decision trees and neural networks to guide the generation of synthetic data samples. These samples are created through adversarial processes and validated by a neural network trained to distinguish between real and synthetic transactions. The proposed framework significantly improves the performance of traditional machine learning models, achieving remarkable accuracy, precision, recall, and F1 scores. Specifically, our method yields an accuracy of 0.9968, with precision, recall, and F1 scores all exceeding 0.995. This superior performance is a result of effectively addressing the class imbalance in the dataset, leveraging advanced sampling techniques, and employing robust machine learning classifiers. By enhancing the identification of fraudulent activities, our approach provides a substantial improvement in fraud detection systems for credit card transactions, ultimately offering a more reliable and efficient solution to this critical problem.
Keywords:	Credit Card, Fraud Classification, Fraud Detection Techniques
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	OPTIMIZATION OF PUBLIC SERVICE MODEL WITH LIMITED RESOURCES USING LINEAR PROGRAMMING
Author:	A M H PARDEDE, A FAUZI, Y MAULITA, R J SIMAMORA
Abstract:	In daily life, people may require public services, which in this study refer to healthcare, security, protection, and emergency assistance involving threats to public safety. Public services are necessary in both routine and emergency situations where life-saving interventions must be delivered as quickly as possible, so the problem to be solved in this paper is how to provide appropriate and fast services to people who need services by utilizing limited resources. The goal is to provide timely services with limited resources from hospitals, police departments, fire departments, and the Regional Disaster Management Agency (RDMA/BPBD). When delivering public services, medical staff (me), police (po), firefighters (fi), and BPBD (bn) depart from their respective locations such as hospitals (ime), police stations (ipo), fire stations (ifi), and BPBD offices (ibn) to the victims' locations (j) to provide services. The model developed in this research addresses challenges by delivering intelligent services as early as possible, reducing fatalities caused by delays, minimizing costs, shortening service times, maximizing resource utilization, and achieving the maximum value from the model's objective function. The information obtained includes the type of solver used—either "B-and-B," "Global," or "Multi-start," depending on the specific solver employed. The objective value was 169.0000, with an objective bound of 169.0000, zero infeasibilities, zero extended solver steps, 20 total solver iterations, and an elapsed runtime of 0.19 seconds. The model is classified as MILP, consisting of 100 total variables, 128 total constraints, and 750 nonzero elements. The designed modeling results have successfully minimized travel costs, service costs, and other costs arising from inaccuracies in patient care delivery.
Keywords:	Emergency; Health; Model; Optimization; Service.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	DYNAMIC SIGN LANGUAGE RECOGNITION AND TRANSLATION THROUGH DEEP LEARNING: A SYSTEMATIC LITERATURE REVIEW
Author:	YESSANE SHRRIE NAGENDHRA RAO, YUAN TING CHONG, REHMAN ULLAH KHAN, CHEE SIONG TEH, MOHAMAD HARDYMAN BARAWI, MOHD SHAHRIZAL SUNAR, JOAN JO JO SIM
Abstract:	Sign language is the communication tool for deaf and hard-of-hearing (DHH) communities all around the world. But it is still difficult to establish proper communication between hearing and DHH individuals. As a result, numerous explorations and investigations that focused on sign language recognition and translation (SLRT) have garnered significant attention from researchers in related fields. This systematic literature review aims to provide a comprehensive study on current trends of state-of-the-art dynamic SLRT models proposed in 85 journal articles found in the Scopus database from 2020 to 2024. Based on the selected articles, this review produced an in-depth analyzation of dynamic SLRT models in terms of their frameworks, deep learning techniques, datasets, pre-processing techniques, and evaluation metrics used. Additionally, this review also highlights both the advancements and ongoing challenges in the domain. Notably, there have been considerable development in isolated and continuous SLRT models, particularly through the combinations of deep learning algorithms such as Convolutional Neural Network, Recurrent Neural Network and Transformer models, with suitable datasets. However, the complexities and challenges of developing robust continuous SLRT models for real-time SLRT persist. This systematic literature review was prepared to serve as a foundational reference that will assist future studies on dynamic SLRT.
Keywords:	Dynamic Sign Language, Sign Language Recognition, Sign Language Translation, Deep Learning, PICOC Criteria
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	PREDICTION AND ASSESSMENT OF SOFTWARE ENGINEERING SKILLSET AMONG COMPUTER SCIENCE STUDENTS USING CONVOLUTIONAL NEURAL NETWORKS THROUGH EXPLAINABLE AI
Author:	JASMIN NIZAR , R SHARMILA, K U JASEENA
Abstract:	The dynamic nature of the software industry necessitates a focus on actual skills alongside academic proficiency. Success in software development projects is often attributed to a well-rounded combination of Soft skills, Life skills, and Technical expertise. The development of skills stands out as the most reliable method for helping students to acquire the essential competencies needed to navigate the complexities of the modern world. Practitioners in software engineering often express concern that graduates lack the necessary preparation for a successful career in the field. Consequently, it has become imperative for educational institutions and employers to accurately assess and predict the skill sets of software engineering students. Achieving a balance among these competencies is vital for excelling in the collaborative and ever-evolving realm of software development endeavours. The prediction and assessment of software engineering skillsets among computer science students are essential to ensuring that graduates are well-equipped for the evolving demands of the professional landscape. This proactive approach aligns academic programmes with industry requirements, allowing educators and institutions to gauge the effectiveness of their teaching methodologies and curriculum. This study aims to provide a comprehensive evaluation and prediction of software engineering skills among computer science students, employing a Convolutional Neural Network (CNN) model. The Convolutional Neural Network model is utilised to predict the presence of Soft, Life, and Technical skills in students. The architecture involves a 1D Convolutional Neural Network for feature extraction from these skill categories, followed by a fully connected network for classification. The model incorporates Explainable Artificial Intelligence through interpretability, ensuring accountability. Local and vision-based explainability are explored using techniques such as Shapley Additive Explanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), and Layer-wise Relevance Propagation (LRP). The performance of the proposed CNN- XAI model is assessed using metrics such as accuracy, precision, recall, F1 score, and the AUC value. Simulation results show that the suggested CNN-XAI model outperforms other basic models in predictive accuracy.
Keywords:	Convolutional Neural Network, Software Engineering, Explainable Artificial Intelligence, SHAP, LIME, LRP.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text

Title:	CONVOLUTIONAL-NEURAL-NETWORKS BASED DETECTION OF THE PNEUMONIA DISEASE –AN EMPIRICAL INVESTIGATION
Author:	PRADEEP M , CH VENKANNA, V.RAVI SEKHARA REDDY, DR.G.SUDHAVANI, B KEERTHI SAMHITHA, KURRA UPENDRA CHOWDARY
Abstract:	Pneumonia is a fatal illness that primarily affects the developed and can occasionally prove to be life-threatening. Early diagnosis of pneumonia has a significant impact on saving many living things. The main focus of this report is the identification and compilation of pneumonia patients based on their chest X-rays. Without any preparation, a convolutional cerebrum network is employed to arrive at the aforementioned conclusion and maintain a remarkable level of precision. When used with X-light emissions, significant learning models automate the framework and guarantee quick, accurate, and comprehensive results. After the image has been processed through a series of convolutional and most limit pooling layers activated by the ReLU incitation work, the class takes place. The neurons in the thick layers are then processed, and finally, the sigmoidal limit activates the final neuron. As the model trains, the precision improves and the adversity decreases. Applying data development prior to model fitting eliminates overfitting. Accordingly, the suggested important learning models to organize the chest X-radiates for the differentiating evidence of pneumonia achieve successful and robust results.
Keywords:	CNN, Pneumonia, ReLU, Empirical Investigation
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2024 -- Vol. 102. No. 21-- 2024
Full Text