		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
September 2025 | Vol. 103 No.17

Title:	AN ARTIFICIAL INTELLIGENCE SERVER DESIGN AND THE ROLE OF THREADS IN PARALLEL PROCESSOR APPLICATIONS
Author:	DURGA PRASAD NOWDU, NAVYA KAILASAM, DIVESH SINGH SAI, B SARITHA, BANDARU SATYA LAKSHMI, CH.SABITHA, DR ARUN KUMAR UNDAMATLA, CH. CHANDRA MOHAN
Abstract:	With the growing demand for high-performance Artificial Intelligence (AI) applications, the design of efficient server architectures has become critical. Modern AI workloads—such as training deep learning models and processing large-scale data—require significant computational resources, which can be effectively met using distributed computing environments composed of high-performance workstations interconnected by fast networks. To ensure usability, such distributed systems must provide a single-system image, abstracting the complexity of the underlying hardware from end users. In building AI applications for these environments, threads offer a more efficient and practical alternative to traditional processes. Threads have lower overhead for creation, context switching, and communication, making them ideal for implementing fine-grained parallelism in multi-core and multi-processor systems. By utilizing threads, AI servers can efficiently distribute computation across multiple processors, improving throughput and responsiveness. This paper presents a prototype AI server design and investigates how multithreading enhances the performance of AI applications running on parallel processing architectures.
Keywords:	Server Design, Threads, Parallel processor, AI Driven, Computing
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	A NOVEL IMAGE-INVARIANT FEATURE EXTRACTION USING SLIDING WINDOW FOR MACHINE LEARNING
Author:	SRAVAN KIRAN VANGIPURAM , RAJESH APPUSAMY
Abstract:	Feature extraction is an important task in building machine learning and deep learning applications. A gray-level cooccurrence matrix, a histogram of oriented gradients, a local binary pattern, principal component analysis, and linear discriminant analysis are some of the feature extraction methods that are used in a lot of research studies. However, these methods are sensitive to image quality, have trouble with non-linear relationships, are difficult to compute, and cannot capture global or contextual information. These limitations often require additional preprocessing or modifications to enhance their performance in practical applications. The goal of this study is to come up with a feature extraction method that works with all kinds of image changes and can pick up both local and global features to make machine learning classifiers work better. To achieve this, we propose in this research a new feature extraction method that is based on the concept of a sliding window to extract local and global image-invariant features. For evaluating the proposed method, we have used the chest X-Ray medical images from the publicly available Novel COVID-19 Chest X-Ray Repository dataset at Kaggle. We conducted experiments using five benchmark feature extraction methods and eight state-of-the-art machine learning classifiers to assess the significance of the proposed feature extraction. For binary classification, the tests indicated that MLP had better accuracy, recall, precision, specificity, and balanced accuracy than other methods (96.25% for accuracy, 96.05% for recall, 92.4% for precision, 96.34% for specificity, and 96.19% for balanced accuracy). The dense MLP neural network, which has two hidden layers with 1024 and 512 neurons each, was able to correctly classify with a 93.98% accuracy, 92.07% recall, 93.21% precision, and 95.32% specificity. It also had a balanced accuracy of 93.69% when it came to multiclass classification.
Keywords:	Sliding window, Feature extraction, Local patterns, Global patterns, Hyperparameters.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ENHANCING BLOCKCHAIN ANONYMITY USING SECURE MACHINE LEARNING APPROACH WITH ONION AND GARLIC ROUTING IN HEALTHCARE
Author:	RAKSHIT KOTHARI , KALPANA JAIN
Abstract:	The research explores an innovative approach to enhancing blockchain anonymity by integrating onion and garlic routing mechanisms with deep learning techniques, specifically Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks. The study addresses growing concerns about privacy and traceability in blockchain transactions by developing a hybrid system that combines the encryption strengths of onion and garlic routing with the predictive capabilities of recurrent neural networks. Our experimental results demonstrate that this integration significantly enhances transaction privacy while maintaining optimal system performance. The proposed model achieved a 94.7% success rate in obscuring transaction origins and destinations, with a 37% improvement in routing efficiency compared to conventional methods. This work provides a promising framework for privacy-focused blockchain applications in secure communication systems and healthcare.
Keywords:	Onion Routing, Garlic Routing, Long-Short Term Memory, Gated Recurrent Unit, Healthcare, Anonymity, Deep Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	GEOSPATIAL INTELLIGENCE AND MACHINE LEARNING FOR FOREST CONSERVATION: AN INTEGRATED APPROACH
Author:	MOHAMED ATIFI , LAHCEN ZIDANE , AZIZ BOUJEDDAINE , SARA RIAHI , HAMID KHALIFI
Abstract:	When thinking about forest degradation, it is tempting to see it simply as the loss of trees. However, the reality is much more complex. Degradation does not only reduce forest cover, it triggers a chain reaction, disrupting ecosystems, accelerating climate change, and affecting the livelihoods of communities that depend on forest resources. Although a substantial body of research exists, we noticed that many studies tend to treat deforestation, wildfires, and illegal logging as separate issues. From our perspective, this fragmented approach often overlooks how these problems are interconnected and how they amplify each other. Traditional assessment methods also present limitations. They usually rely on qualitative observations and have not yet fully integrated the possibilities offered by Geographic Information Systems (GIS) and machine learning. This gap partly explains why risk prediction and the development of targeted conservation strategies remain so challenging. In light of these observations, we propose an integrated framework that combines environmental and socio economic prioritization through the Analytic Hierarchy Process (AHP), spatial analysis using GIS, and predictive modeling based on machine learning techniques. Our main contribution lies in bridging fragmented approaches by offering a unified decision-support method that leverages spatial analysis and predictive modeling for more accurate and actionable forest risk assessment. Rather than replacing existing approaches, our objective is to complement them and offer decision-makers practical tools to anticipate degradation risks more effectively. We believe that bridging the gap between research and action is crucial for achieving meaningful, long term forest conservation outcomes.
Keywords:	Forest Degradation; Climate Impact; Ecosystem; Wildfires; Deforestation; Multi-Criteria Methodology; Analytic Hierarchy Process (AHP); Geographic Information Systems (GIS); Artificial Intelligence (AI); Machine Learning.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	REDESIGNING THE KNOWLEDGE OF ROBOTIC SYSTEM ENGINEERS THROUGH THE MODEL-PARAMETRIC SPACE OF INFORMATION TECHNOLOGY
Author:	OLEKSIY RYKHALSKYY
Abstract:	The relevance of the topic is driven by the rapid development of digital technologies and the growing complexity of designing robotic systems, which requires new approaches to the representation, structuring, and verification of knowledge. The need to integrate heterogeneous models and parameters into a single information environment determines the importance of the study for industrial and scientific applications. The purpose of the study is to investigate methods of representing the knowledge of robotic system designers based on the model-parameter space and to describe the possibilities of their practical use. The object of research is the process of formalizing knowledge and integrating models in the design of complex technical systems. The methodological basis of the study is a systematic approach, semantic modeling methods, interval analysis, and theoretical set operations on knowledge neighborhoods. As a result of the work, the concept of building a model-parameter space (<M,P>) was formed, a classification of basic concepts and categories in the field of design was developed, and formal measures of compatibility, integrity, and completeness of knowledge were proposed. Algorithms for constructing knowledge neighborhoods and integrating models into holistic methodologies with automated consistency checking have been developed. The structure of the tool and software complex to support the processes of model analysis and synthesis in robotic systems is presented. The practical significance of the results lies in the possibility of using the proposed approaches in the development of digital engineering platforms, reducing the risk of design errors, increasing the adaptability of design solutions, and reducing the development time of complex technical systems.
Keywords:	Model, Parameter, System, Analysis, Design, Research, Complex Object, Knowledge Base, Consistency, Heterogeneous Models, Information Technology
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	TECHNOLOGY ORGANIZATION ENVIRONMENT MODEL OF DIGITAL TRANSFORMATION IN PUBLIC ACCOUNTANT PROFESSION FOR GOVERNANCE ENHANCEMENT
Author:	BAMBANG LEO HANDOKO , SARAH JOHANA MAGDALENA SIREGAR , MUHAMMAD FAIZ ATTHALLAH
Abstract:	In today s digital era, digital transformation has expanded beyond industrial sectors and is increasingly penetrating public accounting firms. This transformation is believed to enhance governance by improving efficiency and effectiveness. This study aims to examine the factors influencing digital transformation adoption using the Technology-Organization-Environment (TOE) framework. A quantitative method was employed, and primary data was collected through questionnaires distributed to auditors working at public accounting firms. Data analysis was conducted using Structural Equation Modeling with Partial Least Squares (SEM-PLS) via SmartPLS version 4. The results indicate that technological, organizational, and environmental factors significantly and positively influence digital transformation adoption. Furthermore, digital transformation adoption is found to have a significant and positive effect on governance enhancement.
Keywords:	Accountant, Digital, Transformation, Governance, Enhancement
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	AI-ENABLED SYSTEM FOR IDENTIFYING AND MANAGING INAPPROPRIATE OFFENSIVE CONTENT DETECTION AND MODERATION
Author:	S. SEETHALAKSHMI , Dr. B. BALAKUMAR
Abstract:	This research tackles the important problem of offensive data detection and classification using deep learning methods within the context of social networks. The end goal of this research is to create an artificial intelligence (AI) system that can identify inappropriate data in many forms of media, such as text, audio, and images. The system can detect potentially harmful items by using technologies such as Optical Character Recognition (OCR), Google Text to Speech (GTTS), and Natural Language Processing (NLP). Problems plaguing this area of study at the moment include small datasets and biased model results. Using conventional metrics like Accuracy, Precision, Recall, and F1-measure, the research conducts an exhaustive review to determine the effectiveness of various techniques. According to the results, deep learning models, and the Recurrent Neural Network (RNN) architecture, in particular, are very effective. By facilitating the early detection and prevention of cyberbullying, our study contributes substantial new information towards the goal of creating safer and more inclusive societies. Skillfully extracting data from social media platforms. Our models outperform prior research in identifying and classifying foul language thanks to our use of sophisticated preprocessing techniques and meticulous hyperparameter adjustment.
Keywords:	Offensive data, Artificial Intelligence, Deep Learning, Optical Character Recognition, Google Text to Speech, Natural Language Processing, Recurrent Neural Network
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	SARTIFICIAL BEE COLONY-DRIVEN FEATURE SELECTION EMPOWERING NAIVE BAYES CLASSIFIER FOR SUPERIOR CORONARY ARTERY DISEASE PROGNOSISASE
Author:	JAYAMOL P. JAMES , Dr. GNANAPRIYA S.
Abstract:	The rise in mortality due to Coronary Artery Disease (CAD) underscores the urgent need for accurate, early-stage detection systems. Traditional classification models often struggle with imbalanced datasets, irrelevant features, and limited interpretability, hindering clinical reliability. This study introduces an optimized hybrid diagnostic model Firefly Swarm Optimization-based Decision Tree (FSO-DT)designed to overcome these limitations. The approach leverages the bio-inspired characteristics of firefly swarms to perform robust feature selection, reducing dimensionality and enhancing classification relevance. The refined feature set is subsequently processed through a decision tree classifier, selected for its transparency and clinical interpretability. The proposed model was evaluated using a benchmark CAD dataset, assessing performance through key metrics including accuracy, sensitivity, specificity, and F1-score. Results indicate significant improvements compared to conventional classifiers and unoptimized decision trees. This optimization not only improves classification performance but also ensures model efficiency by minimizing computational overhead. The FSO-DT model delivers an interpretable and scalable solution, capable of aiding clinicians in risk assessment and decision-making processes. The bio-inspired nature of the algorithm ensures adaptability across diverse clinical scenarios, making it a promising tool in medical data analytics. This work contributes to intelligent healthcare systems by integrating evolutionary computation with interpretable learning, fostering advancements in early diagnosis and patient-centered outcomes..
Keywords:	Coronary Artery Disease, Firefly Swarm Optimization, Decision Tree, FeatureSelection, Medical Diagnosis, Classification Accuracy
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	DEEP-LEARNING MODEL FOR BREAST TUMOR IDENTIFICATION USING A HYBRID APPROACH WITH PRE-TRAINED MODELS
Author:	RAMI YOUSEF
Abstract:	Globally, breast cancer is among the most common types of cancer affecting women, and early detection can save lives. In case of breast cancer, early detection improves the chances of a favorable prognosis since prompt treatment can be commenced. Despite the lack of access to specialist physicians, machine learning facilitates early diagnosis of breast cancer. With the rapid advancement of machine learning, especially deep learning, the medical imaging community is increasingly interested in using these methods to improve cancer detection accuracy. Nevertheless, significant deficiencies persist in the existing literature: primary challenges encompass the restricted availability of extensive, balanced, and meticulously annotated datasets; complications regarding model generalizability and external validation; and enduring class imbalance that may result in false negatives and affect clinical reliability. Moreover, there is an absence of established standards and challenges in comparing studies owing to different datasets and evaluation processes. Many deep learning models exhibit robust performance on internal test sets; however, they frequently struggle to sustain accuracy when evaluated on external, heterogeneous data, which raises questions regarding their practical applicability. A limited amount of data is available on diseases. The objective of this study is to present a deep learning model for identifying and classifying breast cancer. Kaggle Open-Access Database breast histopathology images were used to assess the system's performance. Several quantitative criteria are used to evaluate the efficacy of the suggested strategy, including accuracy, precision, recall, and the F1 score. All previous models fail to match the performance of the ensemble model, composed of VGG16, EfficientNetB7, and Xception. As result of the study, the accuracy, precision, recall, and F1-scores for invasive ductal carcinoma (IDC) are all 95.1%, 96.5%, and 95.77% respectively. The results emphasize the strength and dependability of our innovative methodology in automating breast cancer classification, surpassing previous research endeavors, optimizing diagnostic procedures, and ultimately helping to preserve the lives of patients.
Keywords:	MRI Spine Hemangioma Segmentation, U-Net Model, Convolutional Neural Network (CNN), Semantic Segmentation, Precision, Dice Coefficient, Accuracy.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	AN EFFICIENT AND EXTREME LEARNING MACHINE FOR AUTOMATED DIAGNOSIS OF BRAIN TUMOR
Author:	SAYEEDAKHANUM PATHAN, NIKHIL TEJA GURRAM, P. KAVITHA, V NEELIMA, B. SRINIVAS, CHUNDURI LAVANYA A ,BHUPESH DEKA
Abstract:	An efficient automated system is needed to detect the tumor affected and non-effected region of human brain. Though there are traditional models which are already available for the said purpose, But they are facing the problem of excessive segmentation, greater time consumption, and high error rate and over fitting problem.The suggested method with the aid of preprocessing procedures enhances the contrast and quality of the input MRI images. From the pre-processed images, both texture and statistical characteristics are taken out for the subsequent classification process. Further with the aid of the Rapid Sine Cosine Swarm Optimisation method, dimensionality reduction has been accomplished. This has lowered training times and improved model accuracy. To precisely classify healthy and non-healthy images extreme learning algorithm is employed. Finally, to locate the exact tumor affected part of the image auto encoder-based segmentation is used. The proposed model is executed on the BRATS dataset and assessed utilizing a range of performance metrics, including accuracy, recall, F1-score, and precision, which yielded results of 98.93%, 99.21%, 97.67%, and 96.17% respectively, based on the training dataset.
Keywords:	Magnetic Resonance Imaging, Preprocessing, Rapid Sine Cosine Swarm Optimization, Dimensionality Reduction, Auto-Encoder
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ADAPTIVE DEEP REINFORCEMENT LEARNING-DRIVEN TEST CASE PRIORITIZATION FOR HIGHLY VOLATILE SOFTWARE ENVIRONMENTS
Author:	SRINIVASA RAO KONGARANA, ANANDA RAO A, RADHIKA RAJU P
Abstract:	Test execution needs change frequently due to software updates, making fixed test case prioritization unreliable. Unnecessary test runs and delays in finding faults occur because static methods do not adjust when new defects appear. Reordering test cases dynamically based on changing conditions can improve prioritization. A model that updates test rankings using past execution patterns, uncertainty estimation, and retraining triggers is introduced in this study as Adaptive Deep Reinforcement Learning-Driven Test Case Prioritization (ADRL-TCP). Long Short-Term Memory (LSTM) combined with a Double Deep Q-Network (DDQN) stores test effectiveness across multiple cycles. Test cases are ranked based on their likelihood of revealing failures using Bayesian uncertainty estimation. Prioritization is updated without restarting the entire learning process through incremental retraining. In continuous integration environments, the model is tested and compared to Reinforcement Learning for Test Case Prioritization (RL-TCP) and Prioritized Experience Replay Based on Dynamics Priority (PERDP). Finding failures requires fewer test executions, and faults are detected earlier using ADRL-TCP. Past outcomes and recent changes influence test rankings due to the memory-based learning approach. High-risk test cases receive earlier execution because prioritization is adjusted based on uncertainty awareness. Reducing test overhead while maintaining fault detection consistency in continuous testing environments is possible through ADRL-TCP, which adjusts as software evolves. Risk-aware ranking, adaptive retraining, and learning-based memory integration improve test prioritization, as observed in these findings.
Keywords:	Test Case Prioritization, Deep Reinforcement Learning, Continuous Integration, Uncertainty Estimation, Adaptive Learning, Automated Retraining, Software Testing, Fault Detection.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ENHANCING QUALITY OF SERVICE FOR ROUTING IN IOT USING THE PROPOSED DEEP BELIEF LION OPTIMIZATION (DBLO) ALGORITHM
Author:	R.YANITHA, DR.M.LOGAMBAL
Abstract:	Modern communication networks have been completely transformed by the Internet of Things (IoT), which makes it possible for a wide range of devices to exchange data efficiently. But maintaining the best possible Quality of Service (QoS) in IoT networks—especially in cluster-based architectures—remains a major difficulty. In order to improve QoS metrics in IoT routing, this study presents the Deep Belief Lion Optimization (DBLO) algorithm, a novel combination of the Lion Optimization and Deep Belief Networks (DBN). Modern techniques such as Ant Colony Optimization (ACO), Krill Herd Algorithm, and Convolutional Lion Routing Optimization (CLRO), which blends LOA with Convolutional Neural Networks (CNN), are compared to the DBLO algorithm. Critical QoS metrics, including throughput, energy consumption, routing overhead, packet delivery ratio (PDR), and end-to-end delay, are used to gauge performance. Results from experiments show how effective the DBLO algorithm is at maximizing network performance, cutting down on energy use, and guaranteeing dependable data delivery. The suggested method opens the door for more effective and scalable IoT networks by providing a solid solution for QoS improvement in IoT.
Keywords:	Internet of Things (IoT),Quality of Service (QoS), Deep Belief Lion Optimization (DBLO),Lion Optimization Algorithm (LOA),Deep Belief Networks (DBN),Cluster-Based IoT Network, End-to-End Delay, Packet Delivery Ratio (PDR),Routing Overhead, Throughput, Energy Consumption
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	HOW CAN MULTIMODAL DEEP LEARNING DETECT VIOLENCE IN LOW-RESOURCE SINHALA SOCIAL MEDIA POSTS?
Author:	U DIKWATTA , TGI FERNANDO
Abstract:	Social media is an integral aspect of daily life, yet it also serves as an avenue for disseminating violent content. Image-based posts, in particular, serve as powerful vehicles for such content, prompting efforts to automatically detect and combat this issue. While studies predominantly focus on English content, fewer explore languages with limited digital resources, such as Sinhala. Our research pioneers the classification of Sinhala image posts based on both textual and visual features and introduced a foundational work in the context of a low resource language, Sinhala by adapting standard models for this underexplored domain. As the first step, we introduced a new annotated dataset of Sinhala image posts, with extracted textual and visual components. Furthermore, we employed a customized Tesseract OCR and the EAST library for automatic extraction of these components. After pre-processing the textual data, we utilized models, including LSTM, GRU, 1D CNN, and ensemble models combining 1D CNN with GRU, LSTM, BiGRU, and BiLSTM, alongside word2vec embeddings. Additionally, we incorporated XLM-R, a multilingual transformer-based model, to enhance textual classification. For visual analysis, pre-trained models such as ResNet, AlexNet, ViT, and EﬀicientNet were employed. Finally, we employed an SVM-based late fusion multimodal model for classification. Our approach achieved 95% accuracy, surpassing single-modality results. This marks significant progress in addressing violent content in Sinhala image posts on social media. Although architecture employs standard models, adopting these models to Sinhala language and the creation of a baseline dataset marks a significant contribution.
Keywords:	Deep Learning, Multimodal, NLP, Social Media, Violence
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ADVANCING CERVICAL CANCER DETECTION AND EARLY DIAGNOSIS WITH RESNEXT-BASED TRANSFER LEARNING
Author:	G.V.SURESH , GVS RAJU , HARI K. MADIRAJU , MOHAMMED SAJEENA
Abstract:	Cervical cancer is the fourth leading cause of cancer-related deaths among women worldwide, with mortality rates projected to rise significantly by 2030. Despite being highly preventable and treatable when detected in early stage. Previous studies have compared DL-based models with conventional screening methods; however, their performance in detecting early-stage cervical cancer has been poor. This work introduces a new DL-based method for cervical cancer detection using a colposcopy dataset, which involves three simple stages: image enhancement, feature selection, and classification. Image enhancement methods are first used to improve the quality of colposcopy images, thus making feature extraction better, followed by feature selection using simulated annealing to identify the most relevant features for classification. The features are then used to train a ResXNet classifier, achieving an accuracy of 98%. The proposed model is compared with state-of-the-art deep learning models, including VGG16 and ResNet-50, the results demonstrate that the proposed model as achieved improved performance in terms of accuracy and efficiency.
Keywords:	Cervical Cancer, ResNeXt, VGG16, ResNet-50, Transfer Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	AI-ENHANCED ENERGY MANAGEMENT FOR OPTIMIZING RENEWABLE ENERGY INTEGRATION IN SMART CITIES
Author:	T. SUKANYA , CHITTURI SUGUNALATHA , B VINAY KUMAR , A. MUTHUKRISHNAN , GUDISA ROHINI , SALA SUREKHA , ARUNA VIPPARLA
Abstract:	This paper proposes an artificial intelligence-based energy management system implemented on the Internet of Things in smart cities to optimize the amount of renewable energy used in a smart city, reduce costs, and improve stability in the grid. This system combines machine learning methods (LSTM and SVM), Mixed-Integer Linear Programming (MILP) optimization, and reinforcement learning (RL) to predict energy generation and storage, as well as for balancing load in the grid. We validated the results with real data and proved that our model reduced energy costs by 12%, increased the use of renewable energy by 10%, and improved energy balance by 2.3%. Also, grid stability was enhanced by a 66% decrease in failures and 50% in outage periods. Although the system demonstrated potentially successful outcomes, it relies on data quality and computational power. Future efforts will prioritize improving prediction accuracy using up-to-date weather data and expanding the system to encompass larger urban areas. Building systems have an excellent scope for reliable, energy-efficient, and sustainable energy management in smart cities, enabling innovative and eco-friendly urban infrastructure.
Keywords:	AI-Driven Energy Management, Smart Cities, Renewable Energy Optimization, Machine Learning, Grid Stability
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	PREDICTION OF THE INDONESIA STOCK EXCHANGE COMPOSITE WITH TIME-SERIES EXTERNAL AND TECHNICAL FACTORS USING ARTIFICIAL NEURAL NETWORK
Author:	ADITYA EKA PARAMASATYA , VIANY UTAMI TJHIN
Abstract:	Indonesia Stock Exchange Composite is an indicative of the movement of all stocks traded on Indonesia stock exchange. Investor uses this to measure the level of capital gains and benchmark the investment portfolio performance. Accurate prediction of stock market composite remains a challenging task, particularly in emerging markets where external economic factors such as interest rates of central bank, world oil prices, inflation rates, and USD/IDR exchange rate exert significant influence. Existing research in prediction models using artificial neural network often focus on specific stocks and overlook the combined influence of macroeconomic and technical indicators. This research addresses the gap by exploring artificial neural network models to predict the stock exchange composite price by integrates composite technical indicators with external macroeconomic factors. Cross-industry standard process for data mining uses as framework to guide the development. Secondary time-series data period 2019 to 2024 of each factors taken. Empirical results demonstrate that the Neural Net model achieve high predictive accuracy and outperform Deep Learning model. The findings highlight the effectiveness of combining macroeconomic and technical factors in data mining processing with artificial neural network for composite price forecasting and offer practical implications for investors and analysts in emerging markets.
Keywords:	IDX Composite, Stock, Artificial Neural Network, Artificial Intelligence, CRISP-DM
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	NEUROEXPLAINAI: AN EXPLAINABLE AI AND STATISTICAL FRAMEWORK FOR BRAIN TUMOR DIAGNOSIS AND SEVERITY PREDICTION USING MULTIMODAL MRI WITH NEUROFUSIONNET
Author:	V.HARI PRASAD , DR. T. BHASKAR2, SUKANYA LEDALLA , M. VARAPRASAD RAO , A. HARSHAVARDHAN
Abstract:	The correct diagnosis and prediction of malignancy in brain tumors are critical for neuro-oncology, as they directly influence clinical decision-making. Although deep learning models have had notable success in tumor classification and segmentation based on MRI data, most existing approaches are limited in three aspects: building on imaging modalities only, disregarding clinically relevant metadata, and lacking interpretability because of non-integrated explainable AI (XAI). To overcome these limitations, we present NeuroExplainAI, an explainable deep learning framework for a holistic brain tumor diagnosis and grading. We present NeuroFusionNet, a dual task architecture to fuse deep CNN features with hand crafted radiomic descriptors and patient-level clinical metadata in the form of data-attention. This allows for classification (HGG versus LGG) and severity scoring to be performed concurrently. For decisor transparency, both spatial and channel-level explanations are included using Grad CAM++ and SHAP. The model is trained and tested on BraTS 2021 dataset with 98.34% accuracy, 97.73% F1-score and MAE=0.38 for severity prediction. This paper provides novel insights into the clinical interpretability of multimodal fusion and attention-based weighting, in addition to its effect on the predictive performance. Ablation study and comparisons with state-of-the-arts demonstrate the necessity and effectiveness of each component. The incorporation of explainableAI techniques builds trust and improves usability in clinical workflows, making NeuroExplainAI an appealing platform for reliable, interpretable, and individualized brain tumor assessment.
Keywords:	Brain Tumor Diagnosis, Multimodal MRI, Explainable AI, Severity Prediction, Deep Learning Framework
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	AI-BLOCKCHAIN HYBRID SMART CONTRACT MODEL: FRAUD DETECTION AND IMMUTABLE RECORD KEEPING IN INSURANCE
Author:	POORANI MITHILA S , GURUPANDI M , PRIYANKA K , SOLAIPRIYA S
Abstract:	The insurance sector is being transformed through the combination of artificial intelligence (AI) and blockchain technologies. This study proposes the AI-Blockchain Hybrid Smart Contract Model (AIBSCM), which combines AI-based fraud detection with blockchain-based smart contracts to allow for automated insurance claim processing. A synthetic dataset of 1,000 insurance claims was used to train a random forest model, which achieved 92% accuracy on training data; however, real-world testing revealed difficulty in detecting fraudulent claims from under-represented categories. A blockchain simulation was conducted to demonstrate the secure storage and automated execution of claims, with smart contracts giving transparency and immutability. The architecture integrates decentralised oracles, zero-knowledge proofs (ZKPs), federated learning, and a DAO governance mechanism to provide a privacy-conscious, decentralised, and robust solution for the insurance business. Subsequent study will look at real-world deployment and integration with regulations. The integration of these technologies seeks to address traditional insurance systems' issues, such as data privacy concerns and a lack of transparency. By investigating real-world deployment and regulatory compliance, this model has the potential to transform the insurance business by delivering a safe and efficient method for dealing with false claims. This innovative method has the potential to boost client trust while also streamlining insurance company operations. Overall, the combination of blockchain and privacy-conscious technology might result in increased reliability and a transparent insurance sector.
Keywords:	Artificial Intelligence, Blockchain, Smart Contracts, Transparency, Claim Processing.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	OPTIMIZING ENERGY-EFFICIENT ENCRYPTION IN IOT: EXPLORATION OF STREAM AND LIGHTWEIGHT BLOCK CIPHERS
Author:	DR.T.VENGATESH , DR.Y.J.NAZEER AHMED , DR.VITHYA GANESAN , N.KIRUBAKARAN, DR.A.JYOTHI BABU , DR. P.G.SURAJ , R.DEEPAK , D.SUNANTHA
Abstract:	The massive proliferation of the IoT devices has posed intricate problems in the protection of data delivery within limited computational resource. Such environments are usually very demanding to conventional encryption protocols. The paper fills an urgent knowledge gap because it critically compares lightweight block ciphers and stream ciphers to provide an energy-efficient encryption in the IoT-networks. Although the available studies tend to consider performance or security as a key aspect, our study is unique in that it combines the aspects of the throughput, latency and energy consumption of a system to give a multi-dimensional assessment. We can characterize the actual tradeoffs and applicability of the three modes, CTR, OFB, and CFB by deploying them over a real-time testbed. Findings reveal CTR mode offers high efficiency and performance trade off. The study offers practical information on cipher selection which is part of the future development of lightweight encryption approaches with an (energy-constrained environment) focus.
Keywords:	Lightweight Encryption, IoT Security, Stream and Block Ciphers, Energy Efficiency, CTR, OFB, CFB Modes
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	HYBRID DEEP LEARNING APPROACHES FOR CYBERATTACK DETECTION AND PREVENTION IN CRITICAL INFRASTRUCTURE SYSTEMS
Author:	ARUNA VIPPARLA, DR A. MUTHUKRISHNAN, BADDEPAKA PRASAD, KONA KRISHNA PRIYA, SWETHA PENDEM, SHOBANA GORINTLA, ANIL KUMAR PALLIKONDA
Abstract:	This paper proposes a new hybrid deep learning framework to predict, detect, and stop cyber-attacks in critical infrastructure systems. The goal is to increase the resilience of systems like power grids, transportation networks, and water supply systems by combining supervised learning (CNNs), unsupervised anomaly detection (Autoencoders, GANs), and reinforcement learning (DQN). Distinguishing the concentration of CT images through extensive experiments on real patients' data indicates that the proposed model is far superior to the traditional methods in accuracy, precision, and recall by 98.4%, 97.5%, and 98.2%, respectively, and lowers the rate of false positives to 3.5%. Experiment results show the efficiency of our hybrid way of detecting these well-known and new cyber threats with a good detection time (approximately 30ms). This paper helps advance cybersecurity for critical infrastructure by providing a scalable real-time defense that can react to new threats, enhancing the overall system resilience and maintainability.
Keywords:	Deep Learning, Cybersecurity, Critical Infrastructure, Intrusion Detection, Convolutional Neural Networks (CNN)
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	DETECTING SOIL EROSION PATTERNS USING CONVOLUTIONAL NEURAL NETWORKS
Author:	Dr BJD KALYANI , JAYA KRISHNA MODADUGU , Dr CH N SANTOSH KUMAR , Dr SARABU NEELIMA , N M DEEPIKA , RITHIKA NARENDRAN
Abstract:	Environmental sustainability and agricultural output are seriously threatened by soil erosion. This research utilizes deep learning techniques to create an automated system that detects soil erosion from photographs in or-der to solve these issues. Specifically, visual data and structured environ-mental parameters are combined with a Convolutional Neural Network (CNN) to identify soil conditions and forecast the possibility of erosion. Building a CNN model with the purpose of analyzing and categorizing photographs according to soil conditions is the main task of this research. Lay-ers using max-pooling diminish the dimensionality of feature maps and rec-ord the most important characteristics, eliminating the unnecessary information. Fully linked layers then interpret these characteristics to create the final categorization. The model’s output is a binary classification indicating the presence or absence of soil erosion. In addition to image data, the model incorporates structured environmental data such as rainfall and temperature. This data is processed through a separate branch of the network and combined with CNN's output to enhance prediction accuracy. By integrating environmental factors, the model can account for conditions that in-fluence soil erosion, leading to more robust predictions. The CNN model is trained using the Adam optimizer with binary cross-entropy as the loss function. This configuration is suitable for binary classification tasks and helps optimize the model’s performance. The training process includes 20 epochs with a batch size of 8, and class weights are used to address any imbalance in the dataset.
Keywords:	Soil Erosion, Deep Learning, Convolutional Neural Network (CNN), VGG16, Binary Cross-Entropy
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	BIOFUSIONNET: REAL-TIME MULTIMODAL BIOMEDICAL IMAGE AND SIGNAL SYNTHESIS VIA PREDICTIVE ENCODING AND CROSS-RESOLUTION ATTENTION MECHANISMS
Author:	JANARDHAN G, DR.A.UGENDHAR, C NAGESH, RAHUL SURYODAI, HARIPRASAD E, DR. NICHENAMETLA RAJESH, PRAVEEN KULKARNI
Abstract:	The increasing complexity of clinical scenarios necessitates advanced biomedical imaging and sensing frameworks that address real-time challenges, such as multimodal data integration, noise suppression, and accurate decision support. Current solutions often fall short in adapting to diverse imaging environments, real-time constraints, and providing interpretable outputs, limiting their utility in dynamic clinical workflows. To address these limitations, we propose a novel, end-to-end system integrating multiple advanced methods for biomedical imaging and sensing process. Our Adaptive Multimodal Fusion Network (AMFN) combines multimodal imaging data (MRI, CT, Ultrasound, etc.), dynamically weighting modalities using an adaptive fusion module to generate enhanced composite images. Self-supervised learning further reconstructs missing modalities for robust diagnostic accuracy. For real-time insights, the Lightweight Dynamic Segmentation (LDS) module uses a transformer-based attention mechanism in a lightweight neural architecture to segment critical regions of interest with high precision, even on portable edge devices. Noise and artifact suppression are managed by the Contextual Noise Elimination Module (CNEM), which employs convolutional autoencoders and attention-guided filters to enhance image clarity without sacrificing key clinical details. The Predictive Augmented Imaging (PAI) module leverages generative adversarial networks to predict and augment low-quality or incomplete imaging data, ensuring comprehensive results. Finally, the Explainable Imaging Support System (EISS) delivers transparent, interpretable clinical predictions with saliency maps and uncertainty quantification to bolster clinician trust. This integrated framework improves diagnostic accuracy, supports real-time adaptability, and ensures robust usability in resource-limited settings, setting a new benchmark for biomedical imaging in real-world clinical applications
Keywords:	Adaptive Fusion, Biomedical Imaging, Predictive Augmentation, Noise Suppression, Explainable AI, Scenarios
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	A COMPREHENSIVE STUDY OF NON-INVASIVE ANEMIA DETECTION TECHNIQUES: CHALLENGES AND FUTURE DIRECTIONS
Author:	CHAITRA S P , D R RAMESH BABU
Abstract:	Anemia is a nutritional deficiency highly prevalent among under developed and developing nations. Anemia associated with pregnancy poses severe health/ mortality risks to both mother and infant. Early detection of anemia and treatment is the only solution to reduce the mortality risks. In many developing and under developed countries, early detection is challenging due to low reach and higher cost of invasive diagnostics methods. Towards this end, many non-invasive methods have been developed. These methods use various imaging modalities like eye conjunctiva, fingertip, palm, lip mucosa for detection of anemia. These methods can be briefly categorized to conventional and deep learning methods. Though many techniques have been developed in both categories, the area of non-invasive detection still presents many issues in terms of accuracy, false positives etc. Various challenges in image acquisition, feature engineering, classifier design and optimization etc, need to be addressed to achieve higher accuracy and lower false positives. This work does a critical analysis of existing noninvasive techniques for detection and categorization of anemia. Analysis is done in three categories of image acquisition, feature engineering and classification. The aim is to identify the research gaps and present a solution architecture to address those gaps.
Keywords:	Anemia; NonInvasive Methods; Machine Learning; Deep Learning; Palpebral Conjunctiva; Fingernail Bed Analysis;
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	THE IMPACT OF AI GENERATED CONTENT TO INCRESE CUSTOMER ENGAGEMENT ON SOCIAL MEDIA
Author:	SANNAS SALSABILA R , NILO LEGOWO
Abstract:	The use of social media now plays a crucial role for companies in interacting with customers. However, less than 50% of businesses have successfully utilized it effectively. The main issue is that the content is uninteresting and irrelevant, leading to a decreaced in customer engagement. Therefore, this research aims to determine whether AI-Generated Content can enhance customer engagement on social media. This research was conducted by distributing questionnaires to residents of DKI Jakarta, Indonesia, who have purchased products marketed through social media (TikTok and Instagram), where the content generated was AI Generated Content. In this study, a total of 494 respondents were collected. This research conduct using the customer engagement in social media framework (CESM) combined with several other variables that support studies on customer engagement. This study demonstrates that AI-Generated Content (AIGC) significantly enhances customer engagement on social media. Increased social interaction and hedonic motivation effectively address challenges on social media by generating content quickly while maintaining visually appealing quality. Additionally, the improvement in satisfaction, influenced by increased commitment and trust, can trigger active customer interactions, ultimately boosting customer engagement. Meanwhile, positive emotion has a relatively lower impact, as customer responses to AIGC's ability to provide relaxation and humor are often not very strong. Based on the research findings, it is hoped that business owners can address various issues that arise during the marketing process by understanding the potential and limitations in the creation of AI-generated content.
Keywords:	AI-Generated Content, Social Media, Customer Engagement, Customer Interaction, AI Content
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	AI MARKETING FOR REAL-TIME CUSTOMER SEGMENTATION: AUTOENCODER-ENHANCED TIME-EVOLVING CLUSTERING
Author:	IVAN DIRYANA SUDIRMAN , PATAH HERWANTO
Abstract:	Conventional customer segmentation methods often rely on static models that fail to capture dynamic consumer behavior. Consumer behavior is influenced by external factors such as economic shifts, cultural events, and seasonal trends, necessitating an adaptive approach. This study explores AI marketing segmentation as a solution to track and predict consumer state transitions. The research utilizes deep learning-based feature extraction and time-evolving clustering to analyze grocery store transaction data from April to June 2022, covering pre-Ramadan, Ramadan, and post-Ramadan periods in Indonesia. An autoencoder neural network is employed to reduce high-dimensional transaction data into a latent representation, which is then clustered using MiniBatch K-Means to identify evolving customer segments. The study also applies transition analysis to track segment shifts over time. Findings reveal that customer behavior is fluid, with distinct segment transitions occurring in response to Ramadan. The transition analysis highlights that static segmentation models fail to capture these changes. AI-driven marketing segmentation, enables faster tracking of shifting consumer behaviors, allowing businesses to proactively adjust their marketing strategies.
Keywords:	Dynamic Segmentation, Customer Behavior, Clustering, AI Marketing, Deep Learning, Autoencoder
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ACCELERATING PATHOLOGY REPORT DIGITIZATION: A MULTI-ENGINE OCR AND LLM FRAMEWORK FOR HEALTHCARE APPLICATIONS
Author:	NITIN Y. SURYVANSHI, DR. RAKESH J. RAMTEKE
Abstract:	Digitization and structuring of pathology reports are essential in modern healthcare for enhancing patient care, data analytics, and medical research. This study presents a framework called Dual-integrated Text Extraction using Hybrid OCR Engines (DiText-OCR), which leverages multiple OCR tools and domain-specific dictionaries to accurately digitize diverse text types, including printed text and low-quality scans. The extracted text is further processed using Large Language Models (LLMs) for named entity recognition, relationship extraction, and data structuring. The resulting structured data are integrated into healthcare databases and systems, enabling applications in clinical decision support, research, and analytics while ensuring interoperability. Despite its effectiveness, the framework faces challenges, such as handling non-standard report formats, maintaining patient privacy, and addressing the current limitations of OCR and LLM technologies in medical contexts. Future research aims to integrate this system with electronic health records, extend its application to other medical documents, and utilize structured data for advanced research and predictive analytics. By addressing these challenges, the proposed framework has the potential to revolutionize medical data management, ultimately improving patient outcomes, enhancing clinical efficiency, and fostering innovation in healthcare.
Keywords:	Pathology report digitization, DiText-OCR framework, Optical Character Recognition (OCR), Large Language Models (LLMs), healthcare data interoperability, clinical decision support.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	SPARSE CHANNEL ESTIMATION IN MMWAVE MIMO-OFDM USING BI-LSTM WITH CHANNEL SPARSITY REGULARIZATION
Author:	MAMATHA MALVI CHALAVADI , HALUGONA CHANDRAIAH SATEESH KUMAR , VEERENDRA DAKULAGI
Abstract:	The primary objective of millimeter wave (mmWave) Multiple-Input Multiple-Output (MIMO) systems is to effectively estimate the Channel State Information (CSI). Recent research has increasingly utilized the nuclear norm theory to recover a lower-rank structure of channel matrices. Certain sub-optimal solutions to the rank minimization issue arise when addressing nuclear norm-based convex formulations that reduce channel estimation accuracy. To mitigate this, this research developed the Channel Sparsity Regularization – Bidirectional Long Short-Term Memory (CSR-Bi-LSTM), effectively estimates a channel in mmWave MIMO-OFDM. This research incorporates sparsity as a constraint and ensures the method focuses on essential paths, enhancing accuracy and reducing noise. The Bi-LSTM network captures the relationships among OFDM symbols and exploits the spatial correlations between multiple antennas in MIMO. Moreover, it learns the non-linear relationship among pilot signals and channels, which helps to enhance accuracy of channel estimation. The developed algorithm obtained accuracy of 0.0993 for 35 learning rate, 0.990 for 45 learning rate, 0.986 for 55 learning rate, 0.982 for 65 learning rate and 0.976 for 75 learning rate when compared to other conventional techniques.
Keywords:	Channel Estimation, Channel State Information, Millimeter wave, Multiple-Input Multiple-Output and Pilot signals
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	DOMAIN-ADAPTIVE SENTIMENT ANALYSIS THROUGH BAYESIAN NETWORK–LOGISTIC REGRESSION-BASED PROBABILISTIC DEPENDENCY MODELING
Author:	RAMAJAYAM G , Dr. R. VIDYA BANU
Abstract:	Cross-domain sentiment classification presents persistent challenges in opinion mining due to vocabulary drift, contextual ambiguity, and polarity inconsistency across product domains. Traditional classifiers trained on a single domain often fail to generalize, reducing performance when exposed to new, structurally distinct datasets. This research introduces a probabilistically grounded sentiment classification framework—Bayesian Network–Logistic Regression (BN-LR)—designed to address these challenges within multi-domain online review environments. BN-LR integrates a Bayesian Network to model conditional dependencies among sentiment-bearing features, capturing latent inter-feature relationships across syntactic structures. These probabilistic insights are dynamically incorporated into a logistic regression classifier, enabling adaptive feature weighting and uncertainty-aware sentiment inference. As an IT contribution, BN-LR offers a scalable, interpretable, and statistically principled solution suitable for intelligent recommendation engines, feedback analytics, and sentiment-based decision systems across digital platforms. Evaluated on Amazon reviews across four domains, BN-LR consistently delivers high accuracy without requiring domain-specific retraining or external lexicons. The proposed framework enhances real-world information systems by enabling robust cross-domain sentiment generalization, fulfilling a critical need in adaptive text analytics for IT-driven e-commerce intelligence.
Keywords:	Cross-Domain Sentiment Analysis, Bayesian Network, Logistic Regression, Probabilistic Modelling, Online Product Reviews, Domain Adaptation
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	IMPACT OF DIGITAL TRANSFORMATION OF THE HUMAN CAPITAL MANAGEMENT SYSTEM ON STATE REGULATION OF THE NATIONAL ECONOMY DEVELOPMENT
Author:	KOSTIANTYN SHAPOSHNYKOV , ANASTASIIA DUKA , GRYGORIY STARCHENKO , IHOR CHORNODID , DENYS KRYLOV , OLENA SELEZNOVA
Abstract:	The study reveals the role of human capital as a key factor of sustainable economic growth and a tool for state regulation of the national economy in the context of digital transformation. The purpose of the study is to determine the role of human capital in state regulation of the national economy development in Ukraine in the context of digital transformation and to substantiate directions for improving state regulation to increase the management efficiency of this capital. The research used methods of comparative analysis, statistical generalization, structural and logical analysis, graphic modeling, and the inductive-deductive approach. The source of empirical data was indicators of the State Statistics Service of Ukraine and Eurostat for 2019-2023, which made it possible to trace changes in functioning of the human capital system before the start of a full-scale war and under its influence. It is shown that indicators of the educational level, demographic situation, employment, and digital competencies of the population are decisive for formation of the competitive economy. Despite high level of the higher education coverage in Ukraine (28% in 2023), other socio-economic indicators, in particular low GDP per capita, high unemployment and negative natural growth, indicate low efficiency of the human potential realization. Emphasis is placed on demographic challenges, in particular population decline and aging, which threatens stable reproduction of human capital. Comparative analysis with EU countries shows deep gap in economic productivity, which requires systematic modernization of approaches to human capital management. The digital economy is transforming the labor market, increasing the demand for new competencies, while simultaneously displacing traditional forms of employment. Both positive (innovation, new forms of employment, access to the global market) and negative (increasing inequality, displacement of labor) effects of digitalization are identified. The need for digital modernization of the human capital management system based on analytics, forecasting and information technologies is proven. Strategic role of the state in the human capital development is outlined: from investing in education to forming the inclusive digital policy. Priority areas of state regulation are proposed, aimed at increasing digital competencies, adapting the education system, supporting innovative business, regulating new forms of employment and ensuring digital equality.
Keywords:	Human Capital; Digital Transformation; Government Regulation; Employment; Digital Economy; Human Capital Management; Education; Digital Competencies; Investment In Human Capital.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	SYSTEM ARCHITECTURE DESIGN AND ANALYSIS: MESH COMMUNITY OF PRACTICE WITH GAMIFIED SOCIAL COLLABORATION TO STRENGTHEN CYBERSECURITY AWARENESS
Author:	MAYKIN WARASART , PALLOP PIRIYASURAWONG
Abstract:	Cybersecurity awareness remains a major challenge due to the evolving nature of cyber threats and the difficulty of sustaining employees' engagement in traditional training programs. To address this problem, the concept of a mesh community of practice is introduced to be an effective way to increase collaborative awareness and empower the workforce to be better prepared by keeping pace with the latest threats and mitigation practices. In this research, the authors aim to design an architecture of mesh community of practice driven by social collaboration and gamification for cybersecurity awareness enhancement and evaluate the designed architecture. The study began by stating the designed architecture, followed by an evaluation that was based on expert judgment conducted by ten subject matter experts who have related experiences from various organizations. These experts assess and verify the appropriateness of the designed system architecture. The evaluation results indicated that the proposed design (individual component) was rated highest level (mean=4.62, S.D.=0.48), and was also rated highest level (mean=4.58, S.D.=0.69) for entire system, respectively. The findings indicate that the proposed architecture can effectively enhance awareness and engagement, offering a practical model for continuous cybersecurity learning.
Keywords:	Cybersecurity Awareness, Gamification, Mesh Community of Practice, Social Collaboration
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	A SMART AND SECURE CLOUD FRAMEWORK FOR AUTOMATED HEALTHCARE MONITORING THROUGH VOICE PATHOLOGY DETECTION
Author:	B NARASIMHA SWAMY , LAKSHMEELAVANYA ALLURI , SHOBANA GORINTLA, N V S PAVAN KUMAR , V SUMA AVANI , VAHIDUDDIN SHARIFF
Abstract:	As smart cities continue to advance, the demand for secure, automated, and real-time healthcare services is growing to ensure sustainable and high-quality healthcare monitoring. This research introduces a cloud-based framework that integrates smart healthcare devices, environments, and stakeholders within smart cities to enhance the affordability, accessibility, and security of healthcare services. The primary objective is to develop a cloud-based system for real-time voice pathology detection by analyzing voice and electroglottographic (EGG) signals to accurately differentiate between normal and pathological conditions. By leveraging machine learning models such as Gaussian Mixture Models (GMM) for voice disorder classification, healthcare monitoring can be significantly improved, enabling early diagnosis and intervention. Furthermore, this framework aims to enhance the accessibility and scalability of healthcare services by ensuring secure, automated, and remote health monitoring in smart city environments. The proposed system collects voice and EGG signals from internet-connected devices, transmitting them to the cloud for advanced data analysis. A case study on voice pathology detection (VPD) demonstrated the effectiveness of this approach, where local features extracted from voice signals and shape and cepstral features from EGG signals were classified using a GMM, achieving an accuracy of over 93%. The results are then communicated to registered healthcare professionals for definitive diagnosis and appropriate action. By addressing the complex healthcare needs of smart city citizens, this framework provides a secure, scalable, and sustainable solution for real-time healthcare monitoring and decision-making, contributing to the advancement of smart and efficient healthcare services.
Keywords:	Smart Healthcare Monitoring, Voice Pathology Detection; Smart Cities; Cloud-Based Healthcare; Electroglottographic (EGG) Signals; Gaussian Mixture Model (GMM); Real-Time Healthcare Analytics
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ADDRESSING SECURITY CHALLENGES IN DECENTRALIZED ACCESS CONTROL: A BLOCKCHAIN AND ADAPTIVE LEARNING PERSPECTIVE
Author:	CHINTUREENA THINGOM , M. BEULAH VIJI CHRISTIANA , K. BALAJI , T. RAVI , S.B.GOPAL , T VIJAY MUNI , A. THILAGAVATHY , P. JOHN AUGUSTINE , S. NOORAY SASHMI , R G VIDHYA
Abstract:	The Blockchain is also quickly becoming a disruptive technology for providing people with never-before-seen self-governance by virtue of it being secure, transparent, and decentralized in nature. Due to its nature, it has also come to form the basis for various industry applications apart from serving as an early impetus for next-generation paradigms like cloud as well as edge computing. For security concerns data-sharing applications with distributed access control, the current work proposes a new disruptive approach as an adaptive learning model for addressing the above security concerns. Python programming is also implemented for the system validation. A comparison is also made between ALM of the new proposed work with available encrypting algorithms like RSA and AES. Various parameters are also quantified by the research work, and data thus fed is tabulated in fine details. The result demonstrates enhanced security, scalability, as well as computational efficiency by the proposed algorithm.
Keywords:	Cloud Computing; Edge Computing; Decentralized; Blockchain Skillset; Transparent.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	TRANSFER LEARNING STRATEGIES FOR OPTIMISING FACIAL RECOGNITION ACCURACY
Author:	DR. V. V. R. MAHESWARA RAO , SAROJA PATHAPATI , DR. L. K. SURESH KUMAR , S. NARAYANASAMY , AHMED ALKHAYYAT , SAI KRISHNA EDPUGANTI
Abstract:	Human-computer interaction relies heavily on facial emotion recognition (FER), yet creating reliable models is difficult because of inter-subject variability and a lack of personalised data. In this study, we offer a method for improving the accuracy of face recognition with convolutional neural networks (CNNs) that is based on transfer learning. Initially, a generic model is trained using the massive AffectNet dataset; next, it is fine-tuned using the smaller AMIGOS dataset, which contains subject-specific data. The solution is aimed at one-dimensional emotion prediction (valence and arousal) and the Root Mean Square Error (RMSE) of 0.09 and 0.10, respectively. In addition, testing is done for alternative methods of data sampling active and passive to use and to minimize the cost of labelling and accuracy by using greedy sampling and Monte Carlo Dropout Uncertainty Estimation. This approach offers a scalable, efficient solution for personalised FER systems with real-world applications in psychological health monitoring, adaptive interfaces as well as affect-aware computing. The results demonstrate that only 20-30% of labelled personal data is adequate for near-optimal performance.
Keywords:	Facial Recognition, Transfer Learning, Valence, Arousal, Convolution Neural Network, Deep Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	DESIGN OF LOW POWER 9T SRAM CELL WITH EXPANDED NOISE MARGIN
Author:	THAKURENDRA SINGH , VINAY KUMAR TOMAR
Abstract:	SRAM (Static random access memory) cells operate at lower power supply voltages with higher stability got major attention due to their utmost demand in advanced microelectronics devices. In scaled technologies, the power dissipation and stability of SRAM cells become of prime importance due to the aggressive supply voltage scaling. In this context, a novel 9T SRAM memory cell has been proposed with a commendable improvement in both read stability and write stability of the cell along with a reduction in power dissipation. The characteristics parameters of the proposed 9T SRAM cell are compared with state of-the-art designs like six transistors (6T), tunable access transistors 8T (TA8T), D2P11T (data dependent power supply 11T), and 11T SRAM cells. The RSNM (read static noise margin) of the novel proposed 9T SRAM cell is enhanced by 1.95×, 1.92×, 1.87×, 1.90× as of 6T, TA8T, D2P11T and 11T SRAM cells respectively. It occurs as a result by utilization of an isolated read port during read operation. Further, the write ability of the proposed 9T SRAM cell is increased by 1.12×, 1.05×, 1.12×, 1.14× as of 6T, TA8T, D2P11T and 11T SRAM cells respectively. The read power of 9T SRAM cells also gets reduced by 1.48×,1.16×, 1.38×, 1.66× as of 6T, TA8T, D2P11T, 11T SRAM cells. In addition to this, read access time of the proposed 9T cell is reduced by 1.08×, 1.01×, 1.04×. 1.15× as of 6T, TA8T, D2P11T, 11T SRAM cells. In proposed 9T cell, the Ion to Ioff ratio is larger, i.e., 3.5×, 3.45×, 3.54×, as compared to 6T, TA8T, 11T SRAM cells. It shows the applicability of a 9T SRAM cell for a large density array. The simulation work has been performed with Cadence Virtuoso at the 45nm technology node.
Keywords:	Microelectronics, Low power VLSI design,SRAM, Power consumption, stability.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	MULTI-OBJECTIVE OPTIMAL DG PLACEMENT AND SIZING IN DISTRIBUTION SYSTEMS USING LSF-MALO: A STUDY ON VARIOUS DG CONFIGURATIONS AND LOAD
Author:	NUR ATIQAH ABDUL RAHMAN , ZULKIFFLI ABDUL HAMID , NUR ASHIDA SALIM , ISMAIL MUSIRIN
Abstract:	This study investigates the optimization of Distributed Generation (DG) placement and sizing in power distribution systems using the Loss Sensitivity Factor-Mutated Ant Lion Optimizer (LSF-MALO) algorithm. The goal is to minimize real power losses, enhance voltage stability, and reduce economic losses by strategically deploying DG units, such as photovoltaic (PV), diesel, and wind turbines, in distribution networks. The LSF-MALO algorithm’s effectiveness is evaluated using the IEEE 69-bus system, under varying load scenarios and DG configurations. The results are compared with other optimization methods (LSF-Analytical, EP, and MALO), showing that LSF-MALO consistently outperforms them, particularly in multi-type DG configurations like PV-Diesel-Wind and PV-PV-Wind. The findings demonstrate that LSF-MALO not only achieves significant reductions in power losses and improvements in voltage profiles but also provides substantial cost savings, highlighting its potential for improving the efficiency and sustainability of modern power distribution systems. The study underscores the importance of LSF-MALO as a robust tool for optimizing DG integration in the context of decentralized, eco-friendly energy systems.
Keywords:	Distributed Generation (DG), Loss Sensitivity Factor-Mutated Ant Lion Optimizer (LSF-MALO), Distribution Network, Multi-Objective, Voltage Stability, Optimization Technique
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	AUTOMATED DETECTION AND SEVERITY CLASSIFICATION OF KNEE OSTEOARTHRITIS USING A SCALABLE CONVOLUTIONAL NEURAL NETWORK
Author:	RYAVADI BUCHA RAO, KASUKURTHI VENKATA RAO
Abstract:	Osteoarthritis of the knee is a common degenerative joint condition that lowers quality of life and causes disability, especially in people over 60. It is brought on by the knee joint's degenerating cartilage, which causes bone-on-bone contact, discomfort, stiffness, swelling, and restricted movement. Deep neural networks, specifically “CNNs (Convolutional Neural Networks)”, have shown a lot of promise in medical image processing for the identification and classification of diseases. In order to categorize knee osteoarthritis utilizing X-ray pictures into five groups—Minimum, Healthy, Moderate, Doubtful, and Severe—this research presents SCSNet, a deep learning model. Precision, recall, F1 score, and accuracy had been employed to compare the model's performance to three pre-trained transfer learning models: “VGG-16”, “ResNet-50”, and “Xception”. According to experimental results, SCSNet outperformed the transfer learning models in every metric assessed, achieving higher performance with 98% accuracy.
Keywords:	Knee Osteoarthritis, Deep Learning, Classification, Machine Learning, Scalable Convolutional Neural Network, VGG16.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	BIG DATA AND ARTIFICIAL INTELLIGENCE AS TOOLS FOR OPTIMIZING THE BUSINESS PROCESS MANAGEMENT IN ENTERPRISES
Author:	SVITLANA BELOUSOVA , SVІTLANA PROKHORCHUK , NADIIA BAHAN , OLEKSANDRA TSYRA, YURI CHERNENKO, DENYS TKACH
Abstract:	The aim of the study is to identify relationships between the level of using the artificial intelligence (AI) technologies and Big Data at enterprises, digital intensity of business and economic development of countries, in particular the GDP level per capita. The focus is on how the AI and Big data transform business processes, increasing efficiency, adaptability and competitiveness of enterprises. The study used descriptive statistics, correlation and regression analysis, data visualization, as well as the residual analysis to identify countries that deviate from general trends. The source of data was EU enterprise statistics on the digitalization level and the AI implementation, excluding the financial sector. It was found that there is positive relationship between digital intensity of business and the AI implementation level: countries with higher digital indicators are much more likely to use artificial intelligence technologies. There is also the dependence on the economic development level: countries with higher GDP per capita have a higher percentage of enterprises using the AI. At the same time, countries were identified that demonstrate a higher or lower level of implementation than expected - this indicates availability of other factors (policies, institutions, innovation culture) that affect digital transformation of business. Special attention was paid to the analysis of how the use of Big Data and the AI create new points of management decision-making in business processes. These technologies allow enterprises not only to automate typical operations, but also to carry out more accurate and dynamic management at the stages of planning, forecasting, implementation and monitoring of results. As a result, the management architecture itself changes: new points of control and adjustment appear - for example, in real time it is possible to adjust procurement, logistics or customer service strategies based on predictive analytics. This contributes to flexibility of business models and creates the basis for proactive efficiency management. The comprehensive approach to analyzing countries' readiness for the AI implementation, taking into account both digital and economic characteristics, is proposed, which allows for more accurate assessment of the potential of digital transformation in the business environment. The deeper understanding of new points of intervention in business processes that are opened up using the AI and Big Data. Results of the study can be used to shape state digitalization policies, develop strategies to support business innovation, and stimulate the AI implementation, especially in countries with potential that is currently not fully realized. The described management points can be integrated into enterprise management systems to increase productivity.
Keywords:	Digital Intensity, Artificial Intelligence, Big Data, Business Processes, EU Enterprises, Economic Growth, Digital Transformation, Performance Management, Innovation.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	OPTIMIZING VERTICAL DISTRIBUTION TECHNIQUES FOR HERITAGE MULTIMEDIA DATABASES: CHALLENGES AND SOLUTIONS
Author:	SHYMAA H. MAHMOUD, AHMED E. ABDELRAOUF, NAGWA L. BADR, MOHAMED I. ALI
Abstract:	Distributed Multimedia Database Systems (DMDBSs) face critical challenges of sustaining robust performance, influenced by factors such as data fragmentation, allocation, replication, and site clustering. This research introduces an approach for vertical fragmentation of heritage multimedia databases by leveraging the silhouette analysis method alongside the K-means++ clustering algorithm. The proposed approach offers superior centroid initialization and objective cluster validation, resulting in higher cluster quality and improved data compactness. To evaluate the proposed method, experiments were conducted on a real-world dataset from the Antiquities Museum at the Bibliotheca Alexandrina. Previous approaches using K-means clustering with refinement aim to improve data localization by adjusting attribute clusters. However, such methods are limited by random centroid initialization and lack objective validation of cluster quality. Results show that our approach outperforms both the K-means with refinement method and the Differential Bond Energy Algorithm (DBEA), achieving higher fragmentation quality, better data locality, and lower execution time. These findings highlight the practical relevance of the proposed approach for enhancing the efficiency of multimedia data distribution in DMDBSs.
Keywords:	Distributed Multimedia Database Systems (Dmdbss), Vertical Fragmentation, Feature Scaling, K-Means++ Clustering, Silhouette Analysis, Heritage Multimedia Databases.
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ONTOLOGIES FOR ONCOLOGICAL RADIOLOGY: CHALLENGES AND OPPORTUNITIES
Author:	MARIO DI LONARDO , CARLO SARTIANI
Abstract:	Medical imaging examinations, especially Magnetic Resonance Imaging (MRI), interpreted by radiologists in the form of narrative reports, are used to produce and confirm diagnoses in clinical practice at different levels. Being able to accurately and quickly identify the information scattered in the radiologists’ narratives has the potential to reduce workloads, support clinicians in their decision processes, triage patients to get urgent care or identify and cluster patients for research purposes. This is especially critical within the context of the Tumor Boards, multidisciplinary groups made up of different specialists, who regularly meet to discuss oncological patients potentially needing pre/post-surgery treatments and to make diagnostic and therapeutic decisions for them. Nowadays, it is still difficult to access and analyze radiology reports both effectively and efficiently at scale, due to their unstructured nature, the conciseness and often crypticity of the medical jargon used, and the background knowledge usually required for interpreting them correctly and making high-level correlations and conclusions. Privacy concerns introduce further difficulties and impose the adoption of local tools rather than cloud-based ones, hence preventing the use of popular LLMs. Ontologies represent an important tool for easing the automatic process of medical records and radiology reports. Indeed, they can be used to add structure to otherwise unstructured texts; furthermore, they play a key role in data exchange and sharing, as they enable semantic interoperability. In this review paper we will introduce the problems related to the use of ontologies in the medical domain, with particular emphasis on radiology reports, and discuss the prominent advantages that a systematic use of ontologies would generate.
Keywords:	Ontologies, RDF, Mappings, Annotations, ML
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text

Title:	ADVANCING POST QUANTUM CRYPTOGRAPHY DEVELOPING SECURE ENCRYPTION METHODS RESISTANT TO QUANTUM ATTACKS
Author:	REVATHI TALARI , LEELAVATHY.B , DR. ANANTHAN NAGARAJAN , DR.B. HEMANTHA KUMAR , DR RAGHAVENDER K V, DR. N. NEELIMA , ARUNA VIPPARLA
Abstract:	With the current increase in the rate of quantum computing, traditional cryptographic algorithms face a significant risk of compromise, as they are based on mathematical problems that quantum algorithms can efficiently solve. One solution to this challenge has come up in the form of post-quantum cryptography (PQC). By having a hybrid encrypted solution that combines the two different encryption algorithms, NTRU quantum-resistant and fast AES symmetric-key, with a well-integrated combination presented in this paper, we aim to usher in secure and safe data in the post-quantum era. This mixed regime is the most advanced approach that combines classical and quantum encryption. We compare the performance of the system with that of standalone AES, NTRU, and RSA-based systems, examining aspects such as encryption, decryption, key generation, and key size. Benchmarks indicate that the hybrid system has an overhead, but its performance is passable; encryption times are 60ms and 14000ms for 1 KB and 5000 KB of data, respectively, compared to 12ms for 1 KB and 50ms for 5000 KB by AES and NTRU. The same can be said about decryption, with the hybrid system being more efficient than NTRU but still slower than AES. The NTRU key generation times were depicted as unfavorable compared to those of AES, and this margin was seen to increase as the key sizes of NTRU were varied. Compared to executing the same algorithm on a classical computer, the hybrid system is relatively computationally expensive but offers strong quantum resistance. A hybrid encryption system based on the AES-NTRU is a solution to the quantum computing threat to data security. Later projects will be based around various efforts to optimize the NTRU system, also including hardware acceleration opportunities and an increase in the system utilization in real-life environments.
Keywords:	Hybrid Encryption, Post-Quantum Cryptography, NTRU, AES, Quantum Resistance, Cryptographic Efficiency
Source:	Journal of Theoretical and Applied Information Technology 15^th September 2025 -- Vol. 103. No. 17-- 2025
Full Text