		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
February 2025 | Vol. 102 No.4

Title:	THE IMPACT OF ARTIFICIAL INTELLIGENCE ON THE GLOBAL COMPETITIVENESS OF LABOUR MARKETS
Author:	YURII SIKORSKYI , RITA ZABLOTSKA , MARIIA ZUB , HANNA BRATUS , KAMALA DADASHOVA
Abstract:	The aim of the research is to analytically determine and assess the impact of artificial intelligence (AI) on the global competitiveness of labour markets across the world, the United States, the European Union (EU), and China. It also implies the assessment of the impact of the AI development on the global competitiveness of the labour market of the information technology (IT) industry. The research employed the methods of regression analysis, pairwise correlation analysis, and calculation of the dynamics of changes in the studied indicators. The correlation is 0.87 for the world and 0.88 for China, indicating a strong impact of AI on GDP in these regions. In contrast, the figure is 0.53 for the EU, indicating a weaker relationship. The results of the regression analysis give grounds to state that the tendency to increase unemployment against the background of the AI growth is observed only in the USA and the EU. Therefore, it can be argued that AI is having a noticeable impact on economic performance in the US, China, and the world in general among countries that have not been studied. At the same time, its influence in the EU is less pronounced despite the fact that GDP is the result of the activity of all labour markets. An important area of further research should be the study of the global competitiveness of labour markets in India, the Middle East and Africa, which may open up new opportunities for understanding the dynamics of employment and economic development in these regions.
Keywords:	Competitiveness, Correlation, Investment In AI, Regression, Unemployment Rate In The IT Industry.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	AUTOMATED DISEASE DETECTION IN RICE PLANT LEAVES USING A WALD STATISTICAL PIECEWISE REGRESSIVE EXTREME LEARNING CLASSIFIER
Author:	SUHAILA M P, DR.SHEMALATHA
Abstract:	Agriculture plays a very important role in the Indian economy, global food security, and environmental sustainability. Rice plants are affected by diseases due to various fungi, bacteria, viruses, as well as non-infectious factors. Early plant leaf disease recognition is a crucial part in agriculture to significantly improve crop yield as well as superiority. Conventional methods unable to perform accurate rice leaf diseases without increasing time complexity. Therefore, a novel technique called Wald Statistical Piecewise Regressive Extreme Learning Machine (WSPRELM) is introduced for improving the accuracy of plant leaf disease detection with minimal time. Numbers of rice plant leaf images are gathered as of database. Afterward input image are preprocessed to enhance the image quality. Then the ROI segmentation and feature extraction is performed using Russel–Rao indexive statistical region merging technique. Finally, the leaf image diseases are correctly classified into Healthy, Brown Spot, Hispa, and Leaf Blast using Wald statistical piecewise regression by analyzing extracted feature with ground truth features. Experimental results of proposed WSPRELM technique achieve high accuracy (93.26%), precision (0.942), recall (0.948) and F1-score (0.944) with low disease identification time (132.66ms). These results suggest that WSPRELM has the potential to be a robust solution for rice plant leaf disease detection.
Keywords:	Rice Plant Leaf Disease Detection, Extreme Learning Machine, Russel–Rao Indexive Statistical Region Merging Technique, Wald Statistical Piecewise Regression.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	THE USE OF ELECTRONIC SYSTEMS IN INTERROGATION CONDUCTION IN CRIMINAL PROCEEDINGS
Author:	YELZAVETA KUZMICHOVA-KYSLENKO, ANDRIY SHULHA, IVO SVOBODA, IRYNA DZIUBA, OLENA KALHANOVA
Abstract:	The article is dedicated to the study of the use of electronic systems in interrogation conduction in criminal proceedings. The relevance of the study is stipulated by the increasing demand for the improvement of the criminal process by the implementation of electronic systems during interrogations. The use of electronic systems enables ensuring the collection, documentation, and storage of testimony and evidence, and hearings conduction, which enhances the objectivity of investigative actions, reducing the risk of prejudice and possible abuses. The research aim is to study the influence of electronic systems on the process of interrogation conduction in criminal cases. Their effectiveness, accuracy, and influence on the quality of the received evidence are also analysed. Descriptive method, comparative analysis, qualitative data collection method (Questionnaire-survey), quantitative analysis. The received results confirm that implementation of electronic systems for interrogation conduction enables the reduction of errors, increasing the rate of testimony reliability and process transparency and ensuring protection of the rights of suspects and law enforcement officers. The introduction of innovative electronic systems for interrogation conduction has critical meaning for the increase in the effectiveness of the criminal process. Further integration of these technologies into law enforcement practice is important for the improvement of the justice system. The scientific novelty of the study is in the analysis of electronic systems for interrogation conduction and testimony fixing. These systems are seen as the means for enhancing the accuracy of testimonies and reducing the risk of abuses in law enforcement activity. Further studies perspectives include recommendations for the expansion of electronic systems use in investigation practice, training of investigators and other participants of the process to effectively use electronic systems, etc.
Keywords:	Electronic systems, Criminal proceedings, Innovative technologies, Polygraph, “Electronic court”.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	TOWARD SECURE AUDITING: A STUDY ON AUDITOR READINESS IN CYBERSECURITY IMPLEMENTATION USING EXTENDED UTAUT FRAMEWORKS
Author:	NICHOLAS ZEFANYA ISKANDAR, WILLIAM, KEVIN DENISWARA
Abstract:	This study explores cybersecurity’s crucial role and integration within the audit process in enhancing information and data security in facing rising cyberattacks and data threats using the Extended Unified Theory of Acceptance and Use of Technology (UTAUT). Despite the critical nature of this integration, research models exploring the relationship between cybersecurity and auditing remain insufficiently studied. To address this gap, the research adopts a quantitative method to approach external auditors in public accounting firms in the Greater Jakarta area. Apart from the UTAUT framework, this study has adopted negative inhibitors from the Technological Readiness (TR) model to assess auditor perception on using new technology. The data were collected through questionaries and analyzed using SmartPLS. The findings show that auditors’ behavioral intention to use cybersecurity is strongly influenced by performance expectancy, effort expectancy, facilitating conditions, and insecurity, where social influence and discomfort have no significant influence. By developing an extended UTAUT framework, this research aims to explore the readiness of external auditors to adopt cybersecurity within their audit process to improve audit performance and data security.
Keywords:	Cybersecurity, Audit, UTAUT, Technological Readiness (TR), Data Protection
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	ENHANCED CARDIOVASCULAR DISEASE DIAGNOSIS USING MEL-FREQUENCY CEPSTRAL COEFFICIENTS (MFCC) AND MACHINE LEARNING: A COMPARATIVE ANALYSIS OF MACHINE LEARNING CLASSIFIERS
Author:	ABDULLAH ALTAF, HAIRULNIZAM MAHDIN, AWAIS MAHMOOD, ABDULREHMAN ALTAF
Abstract:	Owing to the fact that cardiovascular diseases (CVDs) are one of the main causes of mortality at the global level, so these diseases must be addressed. This study has approached the reported problem through the signals processing of the heart sounds. In particular, state of the art feature Mel-frequency Cepstral Coefficients (MFCC) has been extracted from the cardiac sound waves. Apart from that, five machine learning classifiers—Bernoulli Na¨ıve Bayes (BernoulliNB), Gaussian Na¨ıve Bayes (GaussianNB), Support Vector Machine (SVM), Random Forest, and k- Nearest Neighbors (kNN)—have been used to extract MFCC features in order to categorize heart sound data. In order to check the robustness of these classifiers, frequently used validation metrics like F1 Score, Accuracy, Precision, Recall, G-mean, and Specificity have been employed. The ensuing results demonstrate that the SVM classifier outperforms all the other classifiers showing the highest accuracy and resilience in the identification of cardiovascular disease. By providing important insights into the unique properties of cardiac sound signals linked to various cardiovascular illnesses, the use of MFCC features improves diagnostic capacities. Apart from that, the proposed non-invasive diagnostic method for cardiovascular diseases yields a possible path towards the early identification and treatment. The findings demonstrate how MFCC data may be utilized to efficiently and precisely identify cardiovascular illnesses using machine learning methods, particularly SVM.
Keywords:	Cardio Vascular Disease (CVDs), MFCC, Machine Learning, Machine Learning Classifiers, Heart Disease, Support Vector Machine (SVM), Heart Sound Signal Analysis
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	AN ENERGY-EFFICIENT AND HIGHLY SECURED STARLING MURMURATION-OPTIMIZED DSR ROUTING PROTOCOL FOR FLYING AD-HOC NETWORKS
Author:	S. NANDHINI , K. S. JEEN MARSELINE
Abstract:	Flying Ad-Hoc Networks (FANETs), consisting of unmanned aerial vehicles (UAVs), are increasingly used in aerial surveillance, disaster response, and environmental monitoring applications. Due to their dynamic topology, limited energy resources, and need for reliable communication these networks pose significant challenges in terms of routing, energy efficiency, and security. To address these challenges, the Starling Murmuration Optimization-based Dynamic Source Routing (SMO-DSR) protocol is proposed, which integrates advanced swarm intelligence principles with secure and energy-efficient routing mechanisms. The SMO-DSR protocol optimizes communication in FANETs by dynamically adapting to network conditions, utilizing the natural behavior of starling flocks for path optimization. A critical feature of the protocol is the integration of RSA encryption during the route discovery process, ensuring that data communication is secure from eavesdropping and unauthorized access. The energy efficiency of the protocol is further enhanced through energy-driven clustering and dynamic power adjustment techniques. SMO-DSR was implemented in the NS-3 network simulator, with extensive simulations showing that the proposed protocol significantly improves energy efficiency and ensures secure communication, outperforming traditional routing protocols in both metrics. The results demonstrate that SMO-DSR is a promising solution for secure, energy-efficient communication in UAV-based systems, making it highly suitable for real-world FANET applications
Keywords:	Dynamic Source Routing, Energy Efficiency, Flying Ad-Hoc Networks, RSA Encryption, Starling Murmuration Optimization
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DLERSE: DEEP LEARNING-ENHANCED RECOMMENDATION SYSTEMS FOR E-COMMERCE USER INTERACTION
Author:	BHAVANI B , Dr. D. HARITHA
Abstract:	With the evolving trends in the e-commerce market, incorporating personalized suggestions for users is a critical component of user strategy for any company looking to expand. Businesses may also improve engagement, contentment, and conversion rates, resulting in faster growth through targeted recommendations. Traditional recommendation systems include content-based filtering and collaborative filtering, but they face challenges such as cold start, data sparsity, and scalability. This particular recommendation system tries to address concerns related to bias or limits of typical recommendation systems, which frequently employ deep learning algorithms. We investigate the processing and extraction of complicated human preferences and behaviors from high-dimensional information using deep learning frameworks such as transformer models, recurrent neural networks (RNNs), and convolutional neural networks (CNNs). By accounting for minute details in user interactions, browsing history, context, and more, deep learning models have shown themselves to be more beneficial than standard models, boosting accuracy over time. Thus, backed by our tests, we concluded that the accurate and pertinent recommendations made by deep learning-based recommendation systems enhanced user engagement. These models create fresh suggestions according to user choices, which improves customer happiness and retention in addition to the model's accuracy. According to the study, in the competitive e-commerce industry, employing state-of-the-art deep learning models can lead to more dynamic and captivating user experiences, increasing sales and keeping clients. In conflict-ridden e-commerce ecosystems, the application of the most recent deep learning models can aid in the development of more responsive and user-friendly interfaces, which may boost interest and customer retention. These findings emphasize that, as deep learning advances, its capacity to transform recommendation systems will further boost e-commerce efficacy.
Keywords:	Deep Learning, CNN, KNN, RNN, AI, Recommendation System, E-commerce.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DESIGN AND IMPLEMENTATION OF ELECTRIC VEHICLE CHARGING STATION INTEGRATED WITH DIFFERENT ENERGY SOURCES USING ANFIS CONTROLLER
Author:	B. MOHAN , M.V. RAMESH , MOTURU SESHU , P. MUTHU KUMAR , MADHU VALAVALA
Abstract:	This paper uses the multi-mode operation of solar energy, diesel generator (DG) set, grid, and battery-dependent Electric Vehicle Charging Station (EVCS) for uninterruptable charging and continuous power for in-house loads. A single voltage source converter (VSC) helps EVCS to work in standalone mode, Grid-tied Mode (GTM), and DG set-tied Mode (DTM). Also, VSC deals with different issues in managing proper power-sharing for electric vehicles (EVs) from various energy sources. In addition, to obtain maximum power of solar energy, for the control and monitoring frequency and voltage of the alternator, harmonic current reduction of nonlinear load demands, and required reactive power reduction of the proposed scheme ANFIS (Adaptive Neuro-Fuzzy Inference System) controller is presented. EVCS control strategy is designed to take power from solar energy and a battery. If solar energy and battery fails to supply the EVCS, then it receives electrical power from the main grid, and at the end, it consumes energy from a DG set. The DG set develops 33% additional power against its maximum capability without violating the maximum current flowing through its winding thus, the size of the DG set is minimized. In this work, an ANFIS controller is enforced in place of the PI regulator to regulate the voltage and reduce the THD. Proposed controller reduced the Total Harmonic Distortion (THD). Finally, it enhanced the overall system's performance by improving the power quality.
Keywords:	Diesel Generator, Solar Energy, Voltage Source Converter, Electric Vehicle Charging Station, ANFIS Controller
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	AN OPTIMIZED ATTENTION-BASED DEEP LEARNING MODEL FOR BLACK GRAM LEAF DISEASE CLASSIFICATION
Author:	G SANGAR , V RAJASEKAR
Abstract:	Black gram, a critical pulse crop that accounts for over 70% of global production, is essential for economic stability and nutritional security. Leaf diseases, including Anthracnose, Powdery Mildew, Leaf Crinkle, and Yellow Mosaic, severely jeopardize its productivity, resulting in substantial crop losses. In order to resolve these obstacles, this investigation suggests an automated deep learning-based solution for the early detection and classification of diseases. The research introduces the Efficient AttentionNET model, which is incorporated with Channel Attention and Spatial Attention mechanisms to improve feature extraction, using the Black Gram Plant Leaf Disease Dataset (BPLD) that includes in-field images. The model was able to effectively acquire critical edge information by utilizing wavelet-transformed samples for data augmentation. The SVM classifier with an RBF kernel demonstrated exceptional performance, achieving a 99.50% F1-score, 99.50% precision, 99.52% recall, and 99.50% accuracy. The proposed model is highly effective in the classification of black gram leaf disease due to the integration of wavelet-based augmentation and attention mechanisms. This innovative approach enhances agricultural disease management by assisting farmers in the reduction of yield losses and the promotion of sustainable farming practices. Consequently, it contributes to global food security and economic resilience.
Keywords:	Black gram leaf disease, SVM Classifier, Channel Attention, Spatial Attention, Wavelet Transform, and EfficientNet
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	FAST MASK REGION AND ENTROPY BASED HISTOGRAM EQUALIZED SEGMENTATION OF RICE PLANT DISEASES
Author:	MR. S. KOKILA , DR. S. ABIRAMI , DR. D. NAGARAJU
Abstract:	The most common use of image processing nowadays is in the process of improving images. Conventional contrast enhancement approaches, which including Histogram Equalization (HE), have shown low performance on a wide range of low contrasted image, and are unable to handle various images automatically. These issues arise as a consequence of manually defining characteristics in order to obtain high contrasting images. In this paper, Fast mask Region and Entropy based Histogram Equalized (FMR-EHE) segmentation is proposed. In this research, Entropy based histogram is combined with Fast Mask Region segmentation to improve the accuracy of detecting and classification of Rice Plant Diseases. The proposed technique achieves many objectives, including keeping contrast, conserving the structural properties of the actual histogram, and adjusting the enhancement level. The simulation outputs demonstrate that the proposed method outperforms the published methods in terms of Mean Squared Error (MSE), Peak Signal-To-Noise (PSNR), and entropy calculation.
Keywords:	FMR-EHE, Histogram Equalization, Mean Squared Error, Rice Plant Diseases.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	ADVANCED AI- MACHINE LEARNING METHODS FOR IOT ENVIRONMENT ATTACK DETECTION USING MOUNTAIN GAZELLE OPTIMIZER WITH OPTIMAL DEEP BELIEF NETWORK
Author:	KRANTHI KUMAR LELLA, SWATHI ALLURI, SRINIVASA RAO CHOPPARAPU, NANDITHA BODDU, JAGADESH B N, N. BALAKRISHNA, SENIGE RAJASEKHAR REDDY, M. PADMA
Abstract:	The increasing sophistication of network attacks, including brute-force intrusions, malware distribution, and phishing, poses severe risks to data security, business operations, and financial stability. Traditional Intrusion Detection Systems (IDS) often struggle with inefficient feature selection, high false positives, and poor scalability in IoT environments. To address these challenges, we propose a novel hybrid IDS framework that integrates the Mountain Gazelle Optimizer (MGO) for feature selection with an Optimal Deep Belief Network (DBN) classifier, fine-tuned using the Hybrid Dragonfly-Whale Optimization Algorithm (HDFOA-WOA). Our approach follows a three-stage process: (1) MGO-based feature selection to enhance classification efficiency, (2) DBN-based attack detection, and (3) HDFOA-WOA for hyperparameter tuning to prevent local optima stagnation and improve model convergence. Using the CICIDS2017 benchmark dataset, we validate our model through extensive simulations and k-fold cross-validation, achieving a 98.9% accuracy, outperforming existing IDS models. Our findings demonstrate significant reductions in false positives, improved detection speed, and enhanced adaptability to evolving cyber threats. The proposed approach contributes to real-world cybersecurity by strengthening intrusion detection in IoT networks, ensuring scalable, efficient, and high-precision attack mitigation strategies. Future research will focus on real-time deployment, lightweight model optimization for edge computing, and explainable AI techniques for increased IDS interpretability and transparency.
Keywords:	Optimal Deep Belief Network; Intrusion Detection Systems; Mountain Gazelle Optimizer; Whale Optimization Algorithm; Hybrid-Strategy-Improved Dragonfly Algorithm.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	ENHANCING E-LEARNING THROUGH STRATEGIC STUDENT SEGMENTATION: INSIGHTS FROM THE OULAD DATABASE
Author:	MOHAMED EL GHALI , ISSAM ATOUF , KAMAL EL GUEMMAT , SAID BROUMI , MOHAMED TALEA
Abstract:	This study delves into the transformative potential of data-driven approaches in e-learning, with a specific focus on segmenting students within the Open University Learning Analytics Dataset (OULAD) to optimize personalized education. By employing advanced clustering methods, specifically K-Nearest Neighbors (KNN) and Hierarchical Clustering, the research identifies distinct student profiles based on their demographic information, academic performance, and engagement metrics. Principal Component Analysis (PCA) reduces data dimensionality while preserving essential features to enhance clustering performance and computational efficiency. The results underscore the transformative role of Hierarchical Clustering, achieving higher Silhouette Scores (up to 0.93) and Dunn Index values (up to 2.10) compared to KNN, significantly when PCA is applied, which also reduced computational time by up to 60%. The analysis identified four distinct student clusters, providing actionable insights into their learning behaviors: high engagement but low performance, consistent engagement with high performance, and erratic engagement patterns with fluctuating results. These findings highlight the potential of clustering-based segmentation for designing tailored interventions, ranging from personalized tutoring to motivational strategies, ensuring that e-learning platforms meet the diverse needs of students. By offering a robust framework for scalable and adaptive learning solutions, this study underscores the transformative role of machine learning in enhancing educational outcomes and fostering more inclusive and effective online learning environments, inspiring optimism about the future of e-learning.
Keywords:	Student Segmentation, E-Learning, Learning Analytics, PCA, Hierarchical Clustering, KNN, OULAD, Educational Personalization, Data-Driven Education
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	BREAKTHROUGH FRUITFLY OPTIMIZATION-BASED LEACH ROUTING PROTOCOL (BFO-LRP) FOR PACKET DELAY MINIMIZATION IN WIRELESS BODY AREA NETWORKING (WBAN)
Author:	S. VEERARATHINAKUMAR , Dr. B.DEVANATHAN
Abstract:	Wireless Body Area Networks (WBANs) play a vital role in healthcare and wearable devices, enabling seamless communication and monitoring of sensors attached to the human body. However, efficient routing in WBANs faces challenges due to dynamic body movement, constrained energy resources, and mobility-induced disruptions. To address these issues, this abstract introduces the “Breakthrough Fruitfly Optimization-based LEACH Routing Protocol (BFO-LRP).” BFO-LRP leverages the fruitfly-inspired optimization algorithm, BFO, and incorporates Nonlinear Hierarchical Decision-Making (NHDM) to achieve coordinated and optimized routing decisions. The proposed BFO-LRP aims to overcome data packet delays, limited energy resources, and other routing challenges in WBANs, enhancing the performance of the LEACH protocol. By employing a hierarchical decision-making approach, BFO-LRP ensures adaptive search space exploration, leading to enhanced convergence and efficient exploitation of promising regions. Extensive simulations are conducted using the ns3 network simulator to evaluate its effectiveness. The results demonstrate that BFO-LRP outperforms conventional routing protocols regarding packet delivery performance, reduced delays, and efficient energy utilization, making it a promising solution for routing optimization in WBANs. This research advances WBAN technology, providing better healthcare and wearable applications support.
Keywords:	Delay, Fruitfly Optimization, NHDM, LEACH, WSN, WBANs
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	NEMLAR CORPUS IMPROVEMENT FOR ARABIC NATURAL LANGUAGE PROCESSING
Author:	AYOUB KADIM , AZZEDDINE LAZREK
Abstract:	Most machine learning approaches in Natural Language Processing rely mainly on corpora. Indeed, various applications based on this approaches require prior learning of statistical models, including the Hidden Markov Model for Part Of Speech Tagging. However, this learning resources must meet some criteria to have a well trained model, and thus more accurate results. On the other hand, we find that the Arabic language - despite its vast use on the internet and in social media - has a limited number of linguistic resources for machine learning, especially corpora with morpho-syntactic annotations. Thus, in this article we will treat the Nemlar corpus, one of the richest annotated linguistic corpora for the Arabic language. The aimed version will have several contributions, especially increasing the rate of recognized words and, subsequently, reducing Out Of Vocabulary words (which represents a major problems in many NLP tasks); as well as fine-grain tagging, by separating the words into their smallest possible sub-units, which will open the way to new applications relying on the granular aspect of Arabic. In this article, we will first present the content of the Nemlar corpus. We will then define some criteria in order to improve its structure and enrich its content. We will also present the different modifications made on the original version, including merging POS tags, separating prefixes and suffixes, creating tags for specific cases, etc. in order to lead to the desired form. Then, we will see the experimentation evaluating the new word recognition rate. At the end, we will talk about the advantages and disadvantages of the resulting version.
Keywords:	Corpus, Nemlar, Part Of Speech Tagging, Natural Language Processing, Arabic Language.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	ENHANCING KIDNEY STONE DETECTION: INTEGRATIVE ANALYSIS OF URINE ATTRIBUTES AND MEDICAL IMAGING
Author:	SIRASAPALLI JOSHUA JOHNSON, Dr. LAKSHMI NAGA JAYAPRADA GAVARRAJU, KOVELAKONDA SAKUNTHALA, Dr GRANDHI PRASUNA, V.S.N. MURTHY, VEERAMOHANA RAO REDDY, MADHAN KUMAR JETTY, Dr.G.S.N MURTHY, SRIPADA V S S LAKSHMI, Dr. SIVA KUMAR PATHURI
Abstract:	Kidney stones are becoming a major global public health concern, causing significant morbidity and a significant financial strain on healthcare systems. MRI and CT scans, as well as urinalysis, are the traditional means of detecting them. Kidney stones may be identified and their composition determined by routine urine analysis, but MRI imaging gives us the most information on the size, shape, and location of the stones inside our bodies. By combining urine analysis, MRI imaging, and deep learning, this study elevates the originality of kidney stone detection to a new level. Stated differently, deep learning refers to a specific model that emphasizes You Only Look Once, as opposed to technologies like SVM and DT that employ conventional machine learning techniques. In terms of speed and accuracy, YOLO's real-time detection capabilities surpass those of SVM and DT, enabling the precise and effective diagnosis of kidney stones. The current approach makes it possible to do either MRI or CT scans, which will ultimately be utilized to ascertain the quantity, size, and associated spatial fields of stones, or urine analysis, if data are readily available. The YOLO technology automatically creates bounding boxes around the stones as they are recognized, giving physicians a clear picture of the issue and supplying information for precise measurement and therapy. The experiment's methodology and findings are explained in the article to demonstrate why YOLO is superior to SVM or DT in kidney stone diagnosis.
Keywords:	Kidney stone detection, Urinary analysis, MRI imaging, Healthcare, Diagnostic modalities)
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DEEP LEARNING-DRIVEN MOSQUITO SPECIES IDENTIFICATION USING YOLOV5 FOR DISEASE MONITORING AND CONTROL
Author:	SHAIK SALMA ASIYA BEGUM, GANNAVARAPU GOKUL MADHAV,SHAIK JAMEER, JAMPANI VENKATA YASWANTH VARMA, CHAPPIDI JOHNWESILY, RUQSAR ZAITOON,SPANDANA MANDE, REHANA BEGUM, POLURU ESWARAIAH
Abstract:	The classification of mosquito species is essential for the monitoring of the transmission of mosquito-borne diseases, including malaria, dengue, and Zika. This investigation introduces a deep learning-based methodology that employs the YOLOv5 framework to accurately identify mosquito species from images. The model obtained an accuracy of 98% after being trained on a dataset that included three species: Aedes, Anopheles, and Culex. The method utilizes CSPDarknet53 for feature extraction and PANet for feature aggregation, followed by YOLOv5 for final classification. This system provides a reliable solution for the automated identification of mosquito species, aiding in the prevention of diseases and the monitoring of real-time conditions. The precision, recall, and F1-score all exceed 97%. Furthermore, the model's ability to rapidly process images is facilitated by the use of YOLOv5, making it appropriate for integration with mobile or edge devices for field-based applications. The rapid identification of mosquito species in a variety of environmental conditions is made possible by the high accuracy and efficiency of this approach, which could potentially contribute health authorities in the implementation of timely and targeted interventions. This framework can also be developed to include additional mosquito species or to deal with future datasets, thereby increasing its relevance in the global control of mosquito-borne diseases.
Keywords:	Aedes, Anopheles, and Culex, YOLOv5, PANet, CSPDarknet53.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	HARNESSING IoT AND DEEP LEARNING FOR SUSTAINABLE WASTE REUSE IN CEMENT FACTORIES
Author:	HUSSIEN MOHSON ABIDE, Dr. FADI HAGE CHEHADE, Dr. ZAID F. MAKKI
Abstract:	In this research paper, the innovative application of Deep Learning (DL), especially deep neural network algorithm, is explored to improve the waste management and recycling strategy in cement factories. As important as the cement industry is in building cities, its production is a major contributor to environmental pollution. Large amounts of gaseous waste, slag and kiln dust negatively affect human health and the environment. Traditional waste management strategies lack efficiency and sustainability, which leads to waste of resources and increased landfills. This study aims to build an effective strategy for recycling and improving by-products using an Artificial Intelligence (AI) algorithm for smart management to preserve the environment and predict the best way to work in cement factories. Controlling the feedback weight of neural network derived from the data comes from the Internet of Things (IoT) as the sensors play the key role in enhancing the results. The study showed through the results that SVM was able to identify the best path for optimal waste reuse and dispose of 30% of waste in recycling. The results proved in terms of reducing CO2 emissions and reducing RMSE on historical data and achieved the accuracy of 95% that improved the strategy. This study sheds light on the possibility of using artificial intelligence algorithms as tools to drive sustainability in the industry, especially the cement industry, and the future research avenues in this direction.
Keywords:	Cement Factory, Deep Learning, Internet of Things, Neural Network, Environment.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	EFFICIENT APPROACH FOR PREDICTING SALES USING SUPERVISED MACHINE LEARNING ALGORITHMS
Author:	ANUSHA CHINTAPANTI , SANDIPAN MAITI
Abstract:	The significant impact in businesses is generally affected by manufacturing, planning, supply chain, marketing, warehousing, logistics, and resource management, usually managed by sales forecasting. Casual forecasting techniques and the correlations between factors are used to anticipate future sales behaviour without relying on historical data and trends. Despite the wide usage in research and application, there are severe drawbacks regarding the forecasting techniques related to classic time series. The sales related to supermarkets, along with association rules, regression techniques, time series algorithms, etc., are estimated by numerous available methods. This paper explains constructing a prediction model based on a supervised machine learning algorithm known as Ada Boost to estimate possible sales for 45 Walmart stores in various locations. It is a great opportunity for researchers to predict sales for Walmart, as it is the largest store existing in the world. The sales will be affected on a periodic basis during an event or holidays. This affect might also extend on a daily basis.
Keywords:	Forecasting, Supervised Machine Learning Algorithms, Unsupervised Machine Learning Algorithms, Time Series, Adaboost Algorithm.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	GENERATIVE ADVERSARIAL NETWORKS FOR CYBER THREAT SIMULATION AND DEFENCE STRATEGIES
Author:	DR. M. CHANDRA SEKHAR, VIJAYA ALUKAPELLY, TIRUGULLA NEELIMA, DR. RAJITHA KOTOJU, DR. VADLAMANI VEERABHADRAM
Abstract:	The security of applications and networks is crucial and must be updated regularly. With ongoing technological innovations, adversaries continuously find new ways to compromise systems, highlighting the need to enhance cybersecurity. Traditional approaches like cryptography and firewalls have created safe coding systems and applications. However, due to the vast amounts of data flowing in today's cloud computing environment, it is essential to develop scalable methods for detecting intrusions. The emergence of artificial intelligence has made it possible to utilize deep learning models to detect cyber threats automatically. Literature suggests that there is a need for a generative adversarial network (GAN)-based deep learning framework to improve the quality of training, thereby enhancing the efficiency of intrusion detection. This paper proposes a GAN-based framework for automatically detecting cyber attacks. We use an improved CGAN model for the empirical study. We introduce an algorithm known as Learning-Based Cyber Attack Detection (LB-CAD), which leverages the enhanced CGAN model with the improved VGG16 model to optimize performance in defending against cyber attacks. Our empirical study, using a benchmark known as RT-IoT2022, revealed that the proposed method outperforms many existing approaches, achieving an accuracy of 97.62%. Therefore, the proposed framework can be integrated with existing applications to complement traditional security measures in a scalable manner.
Keywords:	Cybersecurity, Artificial Intelligence, Deep Learning, Cyber Defense Strategies, Generative Adversarial Network
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	NEW FRAMEWORK FOR IMPROVING PRODUCTION EFFICIENCY BY INTEGRATING DEGREE OF CONTRIBUTION AND COST DRIVER FOR THERMAL PRINTER
Author:	NOR SUHADAH RAZALI , MOHD YAZID ABU , NURUL HAZIYANI ARIS , NUR AISYAH MADHIAH HALIM , NUR NAJMIYAH JAAFAR , AHMAD SHAHRIZAN ABDUL GHANI , EMELIA SARI , FAIZIR RAMLIE , WAN ZUKI AZMAN WAN MUHAMAD , NOLIA HARUDIN
Abstract:	Production efficiency is a critical determinant of growth and competitiveness in assessing the success of Malaysia's manufacturing sector. The daily and weekly production reports including information such as working time, output, production efficiency and machine utilization. However, there is no information related to the parameters contributed to the production activities. In meantime, to address production costs, industrial practitioners apply activity-based costing (ABC) to determine the cost per unit of finished products. Obviously, the existing method presents challenges in accurately determining capacity utilization and unused capacity. Regrettably, the quality and costing tools often operate independently, thus the impact of factors to the industrial capacity is less appreciated. This research aims to develop a framework that integrates the degree of contribution and cost driver in production environment. MTS is employed to predict and diagnose system performance using multivariate data for quantitative decision-making. TDABC is utilized as a costing model, enabling companies to allocate costs by calculating the time spent on activities. Data collection is involved 25 workstations, 51 parameters, and 59 activities. As a result, in April 2023, the normal sample has the average MD of 1.000001, while the abnormal sample has the average MD of 53.401398. Increasing the number of parameters which are exceed the normal range will increase the MD value. There are 34% parameters are classified in positive degree of contribution, whereas 66% parameters are classified in negative degree of contribution. For the sub-activity of prepare printing inspection equipment has -22,757.63 minutes and MYR -4,323.95 of unused capacity of time and cost respectively. It was found that there are three types of unused capacity have been identified such as Type I which is the workstation is over-utilized, Type II which is the workstation is small-utilized, and Type III which is the workstation is largely-utilized the resources and cost of apportionment. The proposed framework is great because the degree of contribution reflected the increment or decrement to the cost driver in high production complexity for better product cost.
Keywords:	MTS, TDABC, Optimization, Integration, Capacity Utilization
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DISSOLVED OXYGEN LEVEL MEASUREMENT IN WATER USING IOT AND MACHINE LEARNING
Author:	K.PHANI RAMA KRISHNA, DR.G.V. PRASANNA ANJANEYULU, ANJANEYULU NAIK.R, DR.B.KEERTHI SAMHITHA, DR.J.RAVINDRANADH, SREEDHAR BHUKYA, T.BALAJI
Abstract:	Agriculture, aquaculture, human consumption, and environmental sustainability all depend heavily on water quality. Aquatic ecosystems and biological processes are impacted by dissolved oxygen (DO) levels, which are a crucial indicator of water quality. Conventional techniques for tracking DO levels are frequently labour-intensive, manual, and have a narrow reach. This study introduces a framework for real-time dissolved oxygen measurement and predictive analysis based on the Internet of Things (IoT) and machine learning (ML). The suggested solution makes use of cloud-based storage for real-time access, IoT sensors for data collecting, and machine learning algorithms for precise forecasting and anomaly detection. Scalability, accuracy, and prompt responses in water quality control are guaranteed by this integrated strategy.
Keywords:	Dissolved Oxygen, Iot, Machine Learning, Water Quality Monitoring, Real-Time Data, Environmental Sustainability
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DEEP LEARNING OPTIMIZED FRAMEWORK FOR DETECTION OF ARRHYTHMIA FROM ECG
Author:	MS.K.SHILPA, DR.T.ADILAKSHMI
Abstract:	Early detection of heart diseases has become the need of hour due to surging heart disease occurrence and mortality across world. Electrocardiogram (ECG) test is the most adopted screening test for heart diseases. Various methods have been proposed to extract various features from ECG and use it for classification of heart diseases, still it is open research area due to the need to provide higher accuracy with lower false positives. Optimizations were proposed in various stages like data acquisition, feature engineering, classification stages for achieving higher accuracy. But existing feature engineering approach can be improved with extracting spatial characteristics across multi modalities and temporal characteristics over longer widow duration. This work proposes an optimized deep learning framework to detect Arrhythmia from ECG by extracting spatial characteristics found from multi modalities and temporal correlation over longer time window. Optimization is done in two areas of feature engineering and model parameter fine tuning to provide higher accuracy and lower false positives. The proposed optimization increased the accuracy by 3.5% compared to classifiers without optimization
Keywords:	Heart Diseases, Arrhythmia, Deep Learning, Optimization, Feature Engineering
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	HOW CAN OPTIMIZED ENSEMBLE LEARNING ENHANCE INTRUSION DETECTION? A FEATURE ENGINEERING AND HYPERPARAMETER TUNING APPROACH
Author:	SAYYADA MUBEEN, HARIKRISHNA KAMATHAM
Abstract:	This explosion of new attack vectors renders traditional signature-based detection strategies inadequate for identifying these emerging threats. Additionally, in various high-dimensional data domains, existing methodologies, whether traditional rule-based systems or single-model machine learning approaches, struggle with imbalanced datasets and complex attack patterns. These constraints further result in detrimental accuracy, improper generalization, and ineffectiveness, requiring the development of robust and practical frameworks for intrusion detection. To tackle the abovementioned matters, this study presents an Optimized Ensemble Learning-based Intrusion Detection (OEL-ID) algorithm, a new structure combining feature engineering, hyperparameter tuning, and ensemble learning. The algorithm uses Recursive Feature Elimination (RFE) to extract relevant features to reduce dimensionality and computational time. Insertion of Hyper-models using Bayesian Optimization for fine-tuning the base models hyperparameters (Decision Tree, Random Forest, ExtraTrees, and XGBoost). An ensemble model is constructed utilizing these classifiers through weighted averaging for a robust detection mechanism. It was tested with two datasets, CIC-IDS2017 and NSL-KDD, respectively, using an accuracy of 97.34% and 97.45%. The results show how far the algorithm can go beyond prior approaches in accurately identifying intrusions. Our goal: Strong cybersecurity with OEL-ID algorithm in high-risk federated network scenarios.
Keywords:	Intrusion Detection System, Ensemble Learning, Hyperparameter Tuning, Feature Engineering, Network Security
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	ATTENTION-ENHANCED DEEP LEARNING FRAMEWORK FOR ACCURATE IMAGE FORGERY DETECTION AND LOCALIZATION
Author:	POCHAMPALLY CHANDRA SEKHAR REDDY, BHOOMPALLY VENKATESH, DR. RAJITHA KOTOJU, SUKANYA LEDALLA, RAYAPATI VENKATA SUDHAKAR
Abstract:	The growing accessibility of digital image editing tools has led to severe concerns in domains like forensic investigations, media validation, and cybersecurity. State-of-the-art image forgery detection approaches tend to have limited generalization capabilities across various forgery types (i.e., splicing, copy-move, and AI-produced manipulations), with the ability to localize tampered areas accurately. Recent studies have shown that these limitations arise from inadequate feature refinement mechanisms and adaptability to real-world scenarios. Overcoming these challenges requires a high-performing framework to identify forged images and localize them at the pixel level. This research presents a unique solution involving deep learning-based detection and localization of image forgery, utilizing spatial and channel attention mechanisms to increase the sensitivity of features to forgery artifacts. In this work, we propose a multi-scale feature fusion framework in a UNet-like encoder-decoder architecture to reconstruct the forgery mask precisely. A combination of binary cross-entropy and dice loss is used to optimize this in terms of pixel-wise classification and regional overlap. DL-IFDL is systematically applied to the DEFACTO dataset in a pipeline of preprocessing, feature extraction, attention-based refinement, and conditional random fields for post-processing. The experimental results show that the proposed framework achieves state-of-the-art performance with IoU 96.5% and Dice 98.1%, compared to the existing best method with IoU 93.2% and Dice 96.4%. These results validate the robustness and accuracy of our approach, demonstrating its effectiveness in detecting and localizing forged regions with high precision. This research provides a scalable and adaptable solution that can be integrated into real-world forensic applications.
Keywords:	Image Forgery Detection, Image Forgery Localization, Deep Learning, Attention Mechanisms, Multi-Scale Feature Fusion
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	AN INTEGRATED RELIABILITY MODEL FOR SERIES-PARALLEL SYSTEMS: OPTIMIZING REDUNDANCY ALLOCATION WITH INTERNAL RATE OF RETURN USING HEURISTIC AND DYNAMIC PROGRAMMING APPROACHES
Author:	RAMADEVI SURAPATI, SRIDHAR AKIRI , BHAVANI KAPU, ARUN KUMAR SARIPALLI, SAI UMA SANKAR MANDAVILLI , SRINIVASA RAO VELAMPUDI
Abstract:	The primary objective of reliability engineering is to guarantee that systems and components carry out their intended duties in a consistent manner over a predetermined amount of time and under predetermined conditions. Within the realm of reliability theory, this model is responsible for optimizing system dependability through the strategic allocation of redundancy, all the while balancing limitations like as cost, weight, and volume in series-parallel configurations. The purpose of this study is to investigate the impact that various constraints, specifically weight, volume, dimensions, and spatial limitations, have on the improvement of system reliability. More specifically, this research focuses on spare components for standard drilling machines, which include mechanical elements such as pulleys and gears that enable motion transmission and load management. Utilizing the Lagrangean multiplier method, a system that has an integrated redundant reliability series-parallel configuration is methodically designed and evaluated. This results in the production of real-valued solutions for critical parameters such as component quantities, component reliability, stage reliability, and overall system reliability. For the purpose of obtaining integer answers, the research utilized the heuristic algorithm method as well as dynamic programming approaches. As a result, the analytical precision and importance of the dependability analysis were significantly improved.
Keywords:	IRR Model, Series-Parallel Configuration, System Reliability, LAM Approach, HAM Approach, DMM Approach
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DIGITALIZATION OF REGIONAL ECONOMIC SYSTEMS IN CONDITIONS OF GLOBALIZATION CHALLENGES
Author:	MARTA DERHALIUK , OLENA AREFIEVA , VIKTORIIA CHOBITOK , OLENA KOSTIUNIK , TETIANA SHCHEPINA , INNA SHOSTAK
Abstract:	Relevance of the study is due to increasing role of digitalization in development of regional economic systems, which are increasingly important in the international market in modern conditions. The purpose of the study is to develop and substantiate the methodological approach to assess digitalization of regional economic systems. Methodological basis of the study is digital paradigm, under which generally defined social goals of sustainable development are achieved directed to ensure equal opportunities for all regional subjects in obtaining knowledge and information, access to high-quality services, expanding opportunities to reveal potential of all economic subjects, including households, increasing their safety of life and creating comfort due to digital technologies in everyday life. In the article, the methodological approach to assess digitalization of regional economic systems is proposed, which includes substantiation of principles on choice of evaluation indicators for sub-indices digitalization of regional economic systems (complexity, reliability, strategicity and accuracy, which provides for validity of further calculations of sub-indices and comprehensive index of digitalization of regional economic systems). The methodological approach involves calculating comprehensive index of digitalization of regional economic systems using logic of system analysis in accordance with the “cause and effect constraints” principle; definitions of sub-indices digitalization of regional economic systems based on the weighted geometric mean using the approach on determination of absolute values of the numerical ratios “more”/”less” when comparing corresponding component of sub-indices through the matrix of numerical pairwise comparisons. The typology of regional economic systems is proposed according to values of the complex digitalization index under the adapted Harrington scale into regions with a high, stable, medium and low level of digitalization. The proposed methodological approach to assess digitalization of regional economic systems is tested on the example of regions of Ukraine.
Keywords:	Digitalization, Digital Economy, Digital Technologies, Digital Paradigm, Regional Economic System, Globalization, Region, Economic Entities.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	UTILIZING INTERACTIVE MOBILE TECHNOLOGY FOR HEALTH EDUCATION: CREATION AND EVALUATION OF A CALORIE AND NUTRITION TRACKING APPLICATION FOR THE MALAYSIAN POPULATION
Author:	SARNI SUHAILA RAHIM , CHAN HOE WAI , SHAHRIL PARUMO , ROSLEEN ABDUL SAMAD , SURIATI KHARTINI JALI
Abstract:	This study investigates the transformative capacity of interactive mobile technology as creative tools for revitalizing health education, highlighting their potential to enhance user understanding and promote greater engagement. This article assesses a detailed calorie and nutrition tracking mobile application designed exclusively for the Malaysian demographic. The main aim of this study was to furnish clients with precise and thorough information regarding their nutritional intake, enabling informed dietary and health decisions. The program comprised multiple components, including user registration, a food database and logging system, goal setting and progress tracking, instructional resources, and a mindfulness evaluation. Contemporary calorie and nutrition tracking programs are predominantly designed for users beyond Malaysia and have a limited database of Malaysian food items, making them less relevant to the local populace. This study focused mostly on Malaysian who wanted to track their calorie consumption for weight loss, muscle development, or weight stabilization. The efficacy of the application as a novel instrument for enhancing dietary control and fostering healthy eating practices was assessed through extensive testing with a representative cohort of Malaysian users. Preliminary findings demonstrate that mobile content significantly improves knowledge and awareness, presenting it as a dynamic, engaging, and accessible medium for disseminating information. This effort utilized mobile technologies to improve food knowledge and health outcomes in Malaysia.
Keywords:	Interactive Mobile Technology, Health Education, Nutrition Tracking Application, Malaysian Dietary Practices, Calorie Management and Wellness
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	ETL-POXGB: A NOVEL CLASSIFICATION FRAMEWORK COMBINING ETL DATA INTEGRATION, ENSEMBLE FEATURE SELECTION, AND PSO OPTIMIZATION
Author:	V. USHA , DR. N. R RAJALAKSHMI
Abstract:	The complicated metabolic illness known as diabetes has multiple causes, including genetics, the environment, and lifestyle choices. It is characterized by persistently elevated blood sugar levels. Therefore, to reduce its harmful repercussions, early detection of diabetes is crucial. The growing integration of Information Technology (IT) in predictive healthcare analytics assists in developing more accurate, scalable disease prediction models. IT enhances the current research by employing optimization methodologies, data preparation methods, and machine learning algorithms to increase the accuracy of diabetes predictions. The current research presents an optimization strategy for improving diabetes prediction by combining multiple feature selection models. This work introduces a novel Ensemble Fisher score Kolmogorov-Smirnov score and Chi-Square (FKCS) model that effectively improves forecast accuracy and efficiency. To test machine learning algorithms for predicting diabetes, diabetic datasets like Pima, Iraq, and Frankfurt were used. These datasets came from different sources and had important clinical characteristics. The findings were analysed using multiple statistical machine-learning measures and a stratified cross-validation approach. Among all classifiers, the highest level of accomplishment was demonstrated by the Extract Transform Load: Particle Swarm Optimization XGBoost (ETL-POXGB), achieving an impressive accuracy percentage of 97.16%. The model was validated using Precision, Recall, F1 Score, ROC AUC, CK, and MCCoeff on the merged dataset. In all aspects of evaluation, superior performance was displayed by our proposed model.
Keywords:	Multimodal Medical Data, Optimization, Integrated Diabetic Datasets, Machine Learning, Cross-Validation, ETL-POXGB.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	BTLBD - A FUSION OF CNN, LSTM, AND BIGRU MODEL WITH IMAGE DATA AUGMENTATION FOR IMPROVED BRAIN TUMOR DIAGNOSIS USING OPTIMIZATION
Author:	YENUMALA SANKARARAO, SYED KHASIM
Abstract:	The pressing need for accurate and timely diagnosis of brain tumors underscores the importance of advancing diagnostic technologies. Traditional methods, though effective to a degree, often fall short in terms of precision, accuracy, and speed, crucial factors in brain tumor identification and treatment. This work acknowledges these limitations by introducing a novel deep learning model that identifies brain tumors within a multiclass classification framework. Existing methodologies in brain tumor detection often struggle with lower precision, accuracy, and recall, alongside higher delay times in diagnosis. These shortcomings can lead to significant impacts on patient outcomes, where early and accurate diagnosis is paramount. The Convolution Normalization Mean Filter (CNMF) filter and generative adversarial networks (GAN) for preprocessing are used in a special way in our proposed model to get around these problems. GANs are particularly beneficial in expanding the size of a dataset. The CNMF Filter is a technique for smoothing and reducing noise in images. Fuzzy saliency maps for segmentation. After segmentation, PCA extracts features from the segmented MRI images. We employ the Horse Head Optimization (HHO) technique to select the most optimal features. Finally, we perform classification using a fusion of CNN, LSTM, and BiGRU models. The effectiveness of our model is evident when tested on a dataset containing four brain tumor classes: meningioma, glioma, pituitary, and no tumor. The results demonstrate a significant improvement over existing methods; the proposed model's accuracy in diagnosis is 97.19%, and its accuracy with augmentation is 98.95%. The impacts of this work are far-reaching. By improving the precision and speed of brain tumor diagnosis, our model not only enhances the prospects for timely and effective treatment but also reduces the emotional and financial burdens on patients and healthcare systems. Furthermore, this study's developed methodology establishes a new standard in medical imaging analysis, potentially opening the door for its use in other intricate diagnostic tasks. This work represents a significant step forward in the intersection of deep learning and medical diagnostics, offering a promising tool in the fight against brain tumors.
Keywords:	Brain Tumor Detection, Deep Learning, Recurrent Convolutional Neural Network, Generative Adversarial Networks, Fuzzy Saliency Maps.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	DESIGN OF A HIGH-PERFORMANCE ECG CLASSIFICATION MODEL USING HYBRID GENETIC MACHINE LEARNING MODEL (GM2LH)
Author:	PRAKASH CHANDRA SAHOO , BINOD KUMAR PATTANAYAK, RAJANI KANTA MOHANTY, BIBHUPRASAD MOHANTY
Abstract:	Categorization of electro-cardiogram (ECG) is a primary task for any heart disease classification application. A wide variety of signal processing models are needed in order to perform this task, this includes pre-filtering, noise removal, extraction of features, selection of features, categorization and temporal analysis. Designing a high efficiency ECG classification model requires development & testing of individual methods, and their integration via machine learning models. During integration of these models, inefficiencies are introduced into the system, which reduces final classification accuracy. These efficiencies include, but are not limited to, signal interface between different blocks, classifier feedback, feature selection sizing inefficiency, etc. In order to remove these drawbacks, a novel hybrid Genetic Machine Learning Model (GM2LH) classifier is proposed in this text. Due to an integrated approach taken by the classifier, it is able to classify datasets taken from Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) with 99.7% accuracy, 99.48% precision, 99.15% recall and 99.2% fMeasure performance. The proposed GM2LH model combines feature extraction, feature selection, categorization, and feedback steps into a single integrated approach, which reduces dependency on decentralized blocks. Comparison with state-of-the-art models showcases superiority of proposed GM2LH model, and confirms its utility for different kind of heart diseases. The proposed GM2LH model both enhances classification accuracy and ensures robust feature selection and extraction, making it a promising solution for real-time ECG analysis and early detection of cardiovascular diseases
Keywords:	Electrocardiogram (ECG), Massachusetts Institute Of Technology-Beth Israel Hospital (MIT-BIH), Arrhythmia, Categorization, Genetic Model, Hybrid, Feature Extraction
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text

Title:	HARNESSING IoT AND DEEP LEARNING FOR SUSTAINABLE WASTE REUSE IN CEMENT FACTORIES
Author:	HUSSIEN MOHSON ABIDE, Dr. FADI HAGE CHEHADE, Dr. ZAID F. MAKKI
Abstract:	In this research paper, the innovative application of Deep Learning (DL), especially deep neural network algorithm, is explored to improve the waste management and recycling strategy in cement factories. As important as the cement industry is in building cities, its production is a major contributor to environmental pollution. Large amounts of gaseous waste, slag and kiln dust negatively affect human health and the environment. Traditional waste management strategies lack efficiency and sustainability, which leads to waste of resources and increased landfills. This study aims to build an effective strategy for recycling and improving by-products using an Artificial Intelligence (AI) algorithm for smart management to preserve the environment and predict the best way to work in cement factories. Controlling the feedback weight of neural network derived from the data comes from the Internet of Things (IoT) as the sensors play the key role in enhancing the results. The study showed through the results that SVM was able to identify the best path for optimal waste reuse and dispose of 30% of waste in recycling. The results proved in terms of reducing CO2 emissions and reducing RMSE on historical data and achieved the accuracy of 95% that improved the strategy. This study sheds light on the possibility of using artificial intelligence algorithms as tools to drive sustainability in the industry, especially the cement industry, and the future research avenues in this direction.
Keywords:	Cement Factory, Deep Learning, Internet of Things, Neural Network, Environment.
Source:	Journal of Theoretical and Applied Information Technology 28^th February 2025 -- Vol. 103. No. 4-- 2025
Full Text