		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
January 2026 | Vol. 104 No.1

Title:	A HYBRID STYLOMETRIC TRANSFORMER EMBEDDING FRAMEWORK FOR ROBUST AUTHORSHIP VERIFICATION AGAINST GENERATIVE AI ADVERSARIES
Author:	DR.P.IMMACULATE REXI JENIFER, DR V MAHESH KUMAR REDDY, DR. DUVVURI ESWARA CHAITANYA, LAVANYA KUMARI PITHANI, DR. P.VENKATESWARA RAO , DR. VAISHALI GUPTA
Abstract:	Authorship verification (AV) remains critical to attribute text provenance and to limit misuse of large language models (LLMs). Existing detectors typically used handcrafted stylometric cues or deep transformer embeddings but lacked robustness to model mimicry and targeted obfuscation and often required large labeled corpora. This work proposes Hybrid Stylometric Transformer Embedding Framework (HSTEF), a hybrid, explainable AV architecture that fuses multi-level stylometric descriptors with cross-attended transformer embeddings to form a compact verification representation. HSTEF integrated engineered lexical, orthographic and syntactic stylometric vectors, a pre-trained transformer encoder fine-tuned with contrastive and calibration objectives, and a cross-modal fusion module that enabled bidirectional style–semantic interaction while preserving interpretability. The method was evaluated on the PAN CLEF 2025 Voight-Kampff corpus containing human and machine authored texts across genres and including deliberate obfuscations. Experiments used one classical baseline (Linear SVM on TF-IDF, SVM) and one hybrid baseline (DistilBERT + stylometric features) under identical preprocessing and metrics (ROC-AUC, F1, Brier, C@1, and computation). HSTEF produced an absolute ROC-AUC increase of ≈0.07 and an absolute F1 increase of ≈0.07 over the hybrid baseline while improving calibration (Brier reduced ≈47%) at a controlled inference cost increase consistent with forensic deployment scenarios. These results indicate that a Hybrid Stylometric Transformer Embedding Framework offers a tractable accuracy and robustness trade-off for AV against modern generative AI adversaries.
Keywords:	Authorship Verification, Stylometry, Transformer Embeddings, LLM detection, PAN CLEF 2025.
DOI:	https://doi.org/10.5281/zenodo.18258308
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	ADNET-AI: AN OPTIMIZED DEEP LEARNING FRAMEWORK FOR MULTI-CLASS ECG ARRHYTHMIA DETECTION IN CLINICAL DECISION SUPPORT SYSTEMS
Author:	RAME NIKHILA, BARSHAMOLLA LAVANYA, BOGA VARASREE, DR. K. PURNACHAND
Abstract:	An irregular pulse is reflected by a cardiac condition called an arrhythmia. ECG is the widely used data for the identification of arrhythmia. Healthcare professionals can understand the abnormalities in heartbeats using ECG data. However, with the emergence of AI, there is a need for a CDSS that can help doctors diagnose arrhythmia. We suggested a DL An optimized framework for automatically identifying arrhythmias in ECG data. Our deep learning framework has mechanisms required for data acquisition, data transformation training, a deep learning classifier, and automatic detection of arrhythmia. We proposed a novel DL architecture called ADNet based on the CNN model. The suggested structure has been set up to take advantage of the enhanced CNN model for efficient arrhythmia detection. We presented the Learning-based Arrhythmia Detection method and Classification (LbADC), exploiting Utilizing arrhythmia, the suggested deep learning models detection performance. Our experimental investigation using the MIT-BIH benchmark dataset shows that the suggested method, LbADC, surpasses several cutting-edge models with the best accuracy of 98.68%. Therefore, the proposed AI-enabled system can be integrated with existing healthcare applications to automatically screen arrhythmias in ECG data.
Keywords:	Arrhythmia Detection, DL, Artificial Intelligence, ECG Data Analytics, Hybrid Deep Learning
DOI:	https://doi.org/10.5281/zenodo.18258491
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	THE ROLE OF COUNSELLING GUIDANCE ON CYBERBULLYING BEHAVIOUR AND ITS INFLUENCE ON STUDENT RESILIENCE: A STUDY OF JUNIOR HIGH SCHOOL STUDENTS IN YOGYAKARTA
Author:	EKO PERIANTO, MUNGIN EDDY WIBOWO, HERU MUGIARSO3, SUGIYO
Abstract:	Bullying in the school environment is still a significant challenge for educational institutions in Indonesia, affecting students' resilience, emotional well-being, confidence, and academic engagement. Victims with low resilience often experience social withdrawal, academic difficulties, and psychological distress. This makes schools, especially counseling guidance teachers, play an important role in creating a healthy school environment. This study examines bullying behavior and its impact on the resilience of junior high school students in Yogyakarta. A descriptive qualitative method was used in this study with data collected through interviews by 140 students and 14 guidance and counseling teachers from 14 schools. The results showed that 77.1% of students had experienced bullying, which led to insecurity, anxiety, lack of confidence, frustration, isolation, and helplessness, all of which were closely linked to resilience. Additionally, cyberbullying has emerged as a growing problem, with some students reporting online harassment, negative comments, and ostracization on social media, often in conjunction with bullying at school. Cyberbullying poses a greater risk due to its persistence, anonymity, and wide audience reach, making it more difficult for victims to escape its effects. To mitigate these impacts, schools should adopt comprehensive intervention strategies, including counseling services for victims and perpetrators, anti-bullying education, Cognitive Behavioral Therapy (CBT), the KiVa Anti-Bullying Program, and digital literacy training. Strengthening school policies, safe reporting systems, and support networks is essential to fostering student resilience, enabling them to overcome bullying experiences and thrive academically and socially.
Keywords:	Counselling Guidance, Cyberbullying, Resilience, Students, Intervention Strategies
DOI:	https://doi.org/10.5281/zenodo.18258611
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	ADDRESSING THE DATA SCARCITY OF TAMIL: A ZERO-SHOT COREFERENCE RESOLUTION APPROACH WITH MURIL
Author:	M DHARINI DEVI, DR S KANNAN
Abstract:	Coreference resolution which establishes whether different kind of expressions in a text relate to the same thing, is an essential part of natural language understanding. Although this task has seen substantial development in high-resource languages, particularly through the use of large annotated datasets and pretrained models development. Languages like Tamil under explored because of the limited annotated datasets and specialized language models. To overcome this, a zero-shot method for coreference resolution in Tamil by (MuRIL,) a multilingual pretrained language model is proposed.The framework adapts span-based neural architectures, using contextual span representations, cosine similarity-based scoring, and unsupervised agglomerative clustering to detect and link coreferent mentions efficiently.To facilitate evaluation, a new manually annotated dataset, TGAP (Tamil GAP-style Coreference Dataset) was developed drawn from diverse text sources. Experimental evaluation compared MuRIL with other multilingual baselines using standard metrics such as MUC, B3, CEAF, and LEA F1 scores. Results shows that MuRIL gives the highest overall F1 score (0.67), performing well with other models underscoring the value of language-specific pretraining. The findings shows that a zero-shot, annotation-free framework can achieve competitive performance for Tamil coreference resolution.
Keywords:	Coreference Resolution, Tamil language, Multilingual language model, Zero Shot Learning, Span based Embedding
DOI:	https://doi.org/10.5281/zenodo.18258689
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	MALARIA DISEASE CLASSIFICATION BASED ON MORPHOLOGICAL OPERATIONS AND PCA ALGORITHM
Author:	Mohammed S. H. Al-Tamimi, HADEEL JABAR, BILAL S. H. ALBAYATI, ABEER N. ABDUL-HAMMED
Abstract:	The constant challenge of malaria as a public health issue still remains extremely prevalent in areas with little or no infrastructure or access to health practitioners. Conventional blood-smear analysis is considered to be the "gold standard" approach to detecting malaria, yet it is extremely complex and requires a high level of expert human resources and is prone to error. To overcome these limitations, in the current study a methodology for automated diagnosis of malaria from blood smear images using machine learning algorithms is proposed. The work uses a Kaggle dataset of 27,560 images, split equally into parasitized and uninfected images. Morphological operations - erosion, gradient, bottom hat - are used to improve image quality and highlight important cellular elements. Subsequently, feature extraction is done through Principal Component Analysis (PCA) and classification is done using shallow neural networks, support-vector machines, random forests and logistic regression. Among these algorithms, the shallow neural network had 100 percent accuracy, recall, precision, and F1-score. The results of this study show that morphological preprocessing in combination with Principle Component Analysis (PCA) and simple classifiers could enable the rapid and reliable diagnosis of malaria in resource constrained settings.
Keywords:	Malaria Disease, Classification, Morphological Operation, Image Processing, PCA.
DOI:	https://doi.org/10.5281/zenodo.18258729
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	IMPROVING E-GOVERNMENT THROUGH SENTIMENT ANALYSIS IN THE CONTEXT OF DEVELOPING COUNTRIES: A MAURITANIA CASE STUDY
Author:	MOHAMED EL MOUSTAPHA CHRIF, EL BENANY MOHAMED MAHMOUD, OMAR EL BEQQALI, MOHAMEDADE FAROUK NANNE
Abstract:	New technologies have given rise to online platforms such as Facebook and X (formerly Twitter), which have become de facto channels of communication, enabling people to express their opinions in writing. Analyzing the user-generated data makes it possible to identify opinions and sentiments on specific topics, thereby providing valuable insights for evidence-based decision-making. Therefore, this study focuses on the HASSANIYA dialect, a low-resource variety of Arabic, and proposes a sentiment classification system based on Natural Language Processing (NLP) to support decision-making. In the first step, A dedicated dataset was collected from Facebook, preprocessed using a combination of a Hassaniya-specific stemmer and an Arabic stemmer. Then, different configurations of feature extraction (n-grams) and weighting (TF-IDF) were applied to construct effective feature representations and determine the best classification model. Support Vector Machines (SVM), Logistic Regression (LR), and Random Forest (RF) were used to classify the HASSANIYA dataset. The proposed approach was designed to analyze comments written in both Modern Standard Arabic (MSA) and the HASSANIYA dialect on Facebook and Twitter. The results show that SVM achieved the best performance and was successfully validated through a real case study involving the classification of comments from an X account (Twitter). Finally, this study introduces a prototype dashboard for visualizing sentiment trends, offering a practical tool to support e-government strategies in Mauritania.
Keywords:	Sentiment Analysis, NLP, Feature Extraction, Decision-Making, e-Government, Social Media, HASSANIYA Dialect, Machine Learning Algorithms
DOI:	https://doi.org/10.5281/zenodo.18258766
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	CLIENT-ADAPTIVE NEURAL NETWORKS: A SERVER-DRIVEN ARCHITECTURE FOR DYNAMIC MODEL PARTITIONING IN EDGE COMPUTING
Author:	Dr.POKKULURI KIRAN SREE, Dr.M.MUNI BABU, Dr.K.KISHORE KUMAR, Dr.SESHAM ANAND, Dr.M.SRIDHAR, Dr.N.C.KOTAIAH, Dr.K.SELVAM, DR.T.VENGATESH
Abstract:	The deployment of deep neural networks (DNNs) on resource-constrained edge devices is a significant challenge in contemporary computing. A core problem is that traditional model partitioning, which splits a DNN at a fixed point between a client and a server, fails to account for the dynamic nature of edge environments, including fluctuating network bandwidth, heterogeneous client hardware, and varying computational loads. This paper proposes Client-Adaptive Neural Networks (CANN), a novel server-driven architecture for dynamic model partitioning. In CANN, a lightweight profiler on the client device continuously monitors resource metrics (CPU, memory, battery) and network latency. These metrics are reported to a central Orchestration Server, which leverages a reinforcement learning (RL) agent to dynamically select the optimal partition point from a set of pre-defined "bottleneck" layers within a DNN. This decision minimizes end-to-end inference latency while respecting client resource constraints. We evaluate our framework on a variety of DNN models (MobileNetV2, ResNet-50) and edge device simulators under realistic network conditions. Our results demonstrate that CANN reduces average inference latency by 41% compared to static client-side, server-side, and fixed-split approaches, and improves energy efficiency by 33% on client devices. We conclude that the CANN architecture provides a robust, adaptive solution that effectively overcomes the limitations of static partitioning, paving the way for more responsive and sustainable edge AI systems.
Keywords:	Edge Computing, Dynamic Model Partitioning, Deep Neural Networks, Reinforcement Learning, Resource Constraint, Inference Latency.
DOI:	https://doi.org/10.5281/zenodo.18259315
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM FOR PREDICTING RELIABILITY IN ULTRA-LARGE-SCALE INTEGRATED CIRCUITS
Author:	MRS. SWARNITA G. KALE, DR. RAHUL G. MAPARI
Abstract:	As Ultra Large-scale Integration (ULSI) technology allows for the assembly of billions of transistors on a single chip; it has emerged as the fundamental component of modern machines and embedded systems. Application-specific integrated circuits, memory devices, and microprocessors are all being revolutionized through this advancement, leading to systems that are smaller, faster, and consume less energy. The long-term performance of ULSI systems has been jeopardized by serious reliability issues triggered by the fast scaling of devices into nanoscale levels. Time-dependent dielectric breakdown (TDDB), hot carrier injection (HCI), and negative bias temperature instability (NBTI) are instances of electrical degradation mechanisms that drastically diminish circuits' operating lifetimes. Nonlinear and dynamic deterioration patterns at the nanoscale are frequently ignored by conventional reliability evaluation techniques that rely on statistical distributions or physics-of-failure models. In order to enhance the dependability of ULSI circuits, the study proposed a predictive modeling framework based on machine learning. Using MOSRA-enabled HSPICE simulations of benchmark circuits under stress, reliability data were acquired. Maximum Likelihood Estimation (MLE) and Least Squares Estimation (LSE) have been applied for statistical preprocessing in order to organize the deterioration data. We trained and assessed the prediction accuracy of several ML algorithms, such as Support Vector Regression (SVR), Random Forest (RF), Artificial Neural Networks (ANN), and Adaptive Neuro-Fuzzy Inference System (ANFIS). The findings reveal that ML-driven models, in particular ANFIS, perform better than conventional reliability prediction techniques, as they can accurately simulate nonlinear behaviors and adapt to various stress circumstances. Through introducing predictive reliability techniques, our effort assists in improving the design of next-generation ULSI systems, extending their lifespan proactively, and detecting faults early.
Keywords:	Degradation, Reliability, ULSI, MOSRA/HSPICE, ANFIS, Optimization.
DOI:	https://doi.org/10.5281/zenodo.18259355
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	SECURE IOT COMMUNICATIONS USING SCRP-DRIVEN DYNAMIC QUASIGROUP CRYPTOGRAPHY
Author:	HAITHAM RIZK FADLALLAH , MARWA HUSSIEN MOHAMED , WALID ALAYASH , JANE JALEEL STEPHAN , SARA SALMAN QASIM , TAREQ ALQABANY , MOSTAFA ALI
Abstract:	Modern IoT systems must solve the challenging task of providing secure and efficient communication, as conventional cryptographic algorithms require high computations and/or memory. The paper describes a new lightweight block encryption algorithm with dynamic Theta Quasigroups, which are generated using the Shift Cyclic Random Permutation method. The novelty of this paper is the design of a dynamic, equation-driven quasigroup generation framework that avoids large lookup tables, minimizing the memory overhead while offering strong security with high efficiency in processing. This work contributes to new knowledge by presenting how quasigroup algebra can be turned into a scalable and energy-efficient cryptographic model, optimized for IoT environments. Key entropy and resistance to key recovery and post-quantum attacks are further improved by introducing a 256-bit master key combined with a 128-bit salt and 25-bit random padding. The robustness and randomness properties of the algorithm are confirmed by a security evaluation using the NIST Statistical Test Suite. Experimental comparisons reveal that the proposed method outperforms AES in encryption speed and memory efficiency, particularly for small and medium data sizes, while being compliant with green computing principles. Beyond the purely technical contribution, this work shall also contribute to the educational and theoretical development of lightweight cryptography by providing an easily understandable mathematical framework for teaching and research in IoT security. In summary, the SCRP-driven TQG algorithm proposed here opens a new direction in the design of secure, efficient, and sustainable encryption solutions for IoT and embedded systems.
Keywords:	SCRP; Dynamic Quasigroup; Secure IoT; Lightweight Cryptography; Post-Quantum Security;
DOI:	https://doi.org/10.5281/zenodo.18259402
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	SEMANTICALLY-GUIDED HYBRID TRAVERSAL FOR COMPLEX QUESTION ANSWERING OVER KNOWLEDGE GRAPHS
Author:	M DHARINI DEVI , DR S KANNAN
Abstract:	Knowledge Graph Question Answering (KG-QA) requires strong and effective navigation of complex graph architectures to obtain precise responses from natural language inquiries. Conventional traversal techniques such as breadth-first search (BFS), depth-first search (DFS) , random walks and transformer based approaches often face efficiency challenges like either computationally expensive, prone to combinatorial explosion, lacking semantic focus and failing to scale when applied to large, intricate RDF graphs. To resolve this, we propose an innovative hybrid traversal framework that combines semantic filtering, embedding-based prioritization, and parallel exploration within the RDFLib environment. Our proposed work gives an innovative hybrid traversal framework that effectively combines semantic filtering, embedding-based prioritization and simultaneous exploration within the RDFLib environment. (i) Semantic filtering provides ontology-based restrictions to methodically remove irrelevant pathways, clarity and concentration. (ii) Embedding guided ranking gives knowledge graph embedding to semantically elevate the prioritization of adjacent nodes improving the relevance of traversal choices and (iii) concurrent exploration speeds up the retrieval procedure by simultaneously examining several top ranked paths and scalability. Unlike previous methods, our framework explicitly balances semantic relevance, graph structure, and multi-hop reasoning, while maintaining interpretable paths for answer explanation. The suggested framework undergoes thorough evaluation using a FB15K-237 benchmark dataset of knowledge graph where results demonstrate that our method achieves better performance over baseline models. Initial results show that this combined method provides accuracy, decreases response time, and scalable answer to elevate the capabilities in KG-QA systems.
Keywords:	Knowledge Graph, Question And Answering, Graph Traversal, Semantic Filtering, Embedding-based Ranking
DOI:	https://doi.org/10.5281/zenodo.18259553
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	PROPULSION AND STABILITY AWARE ROUTING PROTOCOL FOR SUSTAINABLE FLYING AD-HOC NETWORKS (SPARROW-EAR)
Author:	R. MYLSAMY , Dr. M. JAIKUMAR
Abstract:	The limited battery capacity of Unmanned Aerial Vehicles (UAVs) remains a critical barrier to sustaining the operation of Flying Ad-Hoc Networks (FANETs), which enable coordinated aerial missions in surveillance, disaster management, and environmental monitoring. Traditional routing protocols emphasize connectivity and throughput but often neglect the dual challenges of propulsion-dominated energy consumption and mobility-induced link instability. This paper presents Propulsion and Stability-Aware Routing Protocol for Sustainable Flying Ad-Hoc Networks (SPARROW-EAR), also referred to as Sustainable Power-Aware Routing with Resilience and Opportunistic Weighting. The protocol integrates a propulsion-aware energy model, residual flight time estimation, and predictive link stability analysis into its routing process. A Pareto-optimized route discovery mechanism filters out non-viable paths early, while a composite route utility function jointly evaluates energy fairness, reliability, and stability. The opportunistic load-balancing mechanism distributes traffic across multiple stable routes, preventing concentrated relay usage and reducing premature energy depletion among UAVs. Simulation results, supported by controlled UAV testbed trials, show that SPARROW-EAR achieves consistently higher packet delivery ratios, reduces energy imbalance by nearly 40%, and extends the overall network lifetime by more than 30% compared with AODV, EMGR, and LSTDA. These performance gains demonstrate SPARROW-EAR's capability to maintain reliable communication and sustained energy efficiency in dense, high-mobility FANET deployments, positioning the protocol as a technically robust solution for mission-critical aerial networking scenarios.
Keywords:	FANETs, UAVs, Energy-Aware Routing, Propulsion Modeling, Link Stability, and Load Balancing.
DOI:	https://doi.org/10.5281/zenodo.18259678
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	TRANSFORMER-BASED AUTOMATION OF AUTHENTIC PRODUCT REVIEWS FOR SCALABLE SOCIAL PROOF IN E-COMMERCE
Author:	PAVAN GUNDA, THIRUPATHI RAO KOMATI, SUBHADRA KOMPELLA
Abstract:	The rapid growth of e-commerce platforms has increased the pressure on a large quantity of genuine and interesting product reviews to create trust and stimulate buying behaviour. Manual review writing is time-consuming and resource-intensive, particularly when the size of the catalog goes into the thousands. New or niche products have a high likelihood of being discovered and not accepted by consumers due to insufficient reviews. Retailers and brands are in need of a solution to automatically create reviews that can replicate the variety and the believability of authentic customer reviews. An AI-generated review system with the transformer-based GPT-2 model that is capable of generating natural and believable product reviews. The automated system will enhance the e-commerce experience through offering scalable, authentic social proof on all product listings. The idea behind this project is to fill this gap between the existence of products and social proof through the use of sophisticated AI methods. With automation of review writing, businesses will be able to provide sufficient support to all product listings and eventually improve customer experience and increase the number of conversions. The study manages to automate the generation of natural product reviews by combining the newest deep learning and natural language processing technologies with the help of GPT-2. The results provided show that the model can capture imperceptible expressions of product functionality, quality, and satisfaction, all of which are critical in generating authentic reviews. The implications of this strategy for the e-commerce sites seeking scalable solutions to sentiment analysis, content development, and enhancement of customer engagement are significant. Another way to add to the work is the introduction of transformer-based architecture that preconditions further research.
Keywords:	Generative AI, GPT-2, Product Review Generation, Natural Language Processing (NLP), E-Commerce Automation, Deep Learning, Transformer Model, Amazon Reviews Dataset, Text Generation, Fine-Tuning
DOI:	https://doi.org/10.5281/zenodo.18259742
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	DEEP EVOLUTIONARY LEARNING MODEL FOR PRECISION MUTATION ANALYSIS IN MYELODYSPLASTIC AND LEUKEMIC CELLS
Author:	K. SRILAKSHMI , D. VENKATA LAKSHMI
Abstract:	Myelodysplastic Syndromes (MDS) are inherited bone marrow stem cell disorders. Comparing the genetic makeup of healthy and MDS patients may help identify and treat the disease. MDS gene expression data is being used by biological scientists to study coding and non-coding gene activities. Categorization and risk stratification are necessary to make clinical decisions concerning Acute Myeloid Leukemia (AML) patients. This study employs two processed gene expression datasets to identify the most critical genes for effective clustering. This research incorporates a deep learning strategy for categorization from segmented images based on the Visual Geometric Group (VGG16) Convolution Neural Network (CNN) architecture. The solution to this problem is a VGG16 deep learning architecture that uses Particle Swarm Optimization (PSO) for disease classification and accurate analysis of gene mutations. Using the PSO technique to fine-tune VGG16's hyperparameters improves feature extraction efficiency and classification performance. A publicly available dataset of pre-processed bone marrow smear images is used to validate the system. This research proposes a Radiation Associated and Therapy Related Rearrangements and Mutations of Gene Analysis for Myelodysplastic Syndrome and Acute Myeloid Leukemia Classification using PSO optimized VGG16 (RATR-RMGA-MDS-AML-PSO-VGG16) for accurate classification levels. A 99.2% classification accuracy, 99.4% VGG16 feature processing accuracy and 99.4% Mutations of Gene Analysis Accuracy Levels were achieved by the suggested PSO-VGG16 model in the experiments, which is a significant improvement over baseline VGG16 and other traditional machine learning approaches. These results show that PSO-VGG16 integration is a powerful and accurate diagnostic tool for MDS and AML classification, which opens new possibilities for treatment plans.
Keywords:	Myelodysplastic Syndromes, Acute Myeloid Leukemia, Intrahepatic Cholangiocarcinoma, Radiation Associated, Therapy, Mutations, Gene Analysis, Particle Swarm Optimization, Convolution Neural Networks.
DOI:	https://doi.org/10.5281/zenodo.18259777
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	TIME-AWARE MACHINE LEARNING WITH MULTI-SOURCE DATA FOR FIELD-SCALE AGRICULTURAL DECISION SUPPORT
Author:	DR. R. SUDHAKAR, DR.B SIVA LAKSHMI, UPPALAPATI NAGA RATNA KUMARI, DR RIDHI RANI, G SANTHOSH KUMAR, DR. P.VENKATESWARA RAO
Abstract:	Field scale decision support for smallholder farms in southern India requires models that respect both space and time. This study aimed to develop a time aware machine learning pipeline that fuses electromagnetic soil conductivity, laboratory soil analyses, Sentinel-2 multi-date imagery and local weather records to provide actionable management zone delineation and nutrient advisory. We process EM surveys and soil cores with ordinary kriging to create spatial rasters, generate weekly Sentinel-2 indices from 2018 to 2022, and align daily weather with crop phenology. A set of models including Random Forest, LSTM, Temporal Convolutional Network and a spatio-temporal Transformer are trained under leave-one-year-out and parcel based cross validation. The Transformer yields the best predictive performance for yield and nutrient response with mean RMSE = 0.42 t ha^-1 and R² = 0.78 on hold out years, improving RMSE by ~12 percent over the Random Forest baseline. Management zone stability measured by Cohen Kappa and Jaccard indices is highest for kriging + fuzzy k-means zonation and supports a 3-class mapping as the best trade-off between stability and interpretability. Practical decision outputs include per parcel N application bands and short advisories designed for field managers. The pipeline reduces the uncertainty of seasonal advisories and delivers maps and short textual guidance usable by extension services in the southern Indian context.
Keywords:	Time Aware Machine Learning, Multi Source Data Fusion, Kriging, Sentinel-2, EM Conductivity, Southern India, Management Zones, Spatio-Temporal Transformer
DOI:	https://doi.org/10.5281/zenodo.18259816
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	EARLY DETECTION OF UTERINE FIBROIDS ENHANCED BY A DUAL-SCALE TRANSFORMER-CNN HYBRID SEGMENTATION NETWORK (DSTCH-NET)
Author:	MINU INBA SHANTHINI WATSON BENJAMIN, VISUMATHI.J
Abstract:	Accurate and early segmentation of uterine fibroids is necessary for quick evaluation, planning of treatment, and better results for patients. However, current segmentation techniques often struggle to achieve reliable results due to the complex morphology, irregular boundaries, low contrast, and noise present in MRI and ultrasound images. This research proposes a solution to these challenges by proposing an innovative Dual-Scale Transformer-CNN Hybrid Segmentation Network (DSTCH-Net) that uses the strengths of both transformer and convolutional architectures. The Fine-Scale Transformer Branch captures the global context and long-range dependencies, while the Coarse-Scale CNN Branch focuses on detailed local feature extraction and texture analysis. A Contextual Cross-Attention Fusion Module (CCAFM) integrates essential features from both branches, and an Adaptive Boundary Refinement Module (ABRM) further enhances edge precision. Experimental analysis shows that DSTCH-Net can be evaluated at 99.2% segmentation accuracy at a loss of 0.07, which is better than the state-of-the-art models. These findings confirm that the suggested method can be used to effectively overcome the constraints of traditional methods and offer a strong framework for accurate and early detection of uterine fibroids in clinical practice
Keywords:	Uterine Fibroids, Medical Image Segmentation, Deep Learning, Dual-Scale Segmentation, Hybrid Architecture, Transformer
DOI:	https://doi.org/10.5281/zenodo.18259934
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	AN ADAPTIVE HASH DRIVEN ACCESS CONTROL MODEL FOR ENHANCED PATIENT DATA SECURITY IN HEALTHCARE
Author:	K.MAITHILI , S.AMUTHA
Abstract:	With the growing volume of health information it has become common practice to protect the patient identity while maintaining convenient access to the data. Due to varying flow of cyber security threats, traditional solutions never manage to get flexible access to data without compromising with overflow of data. To overcome these challenges focusing on patient data protection, in this paper, we propose a new Hybrid Integrated Hashing approach entitled "Dynamic Adaptive Hash-Block Access Control (DAHBAC) framework" using blockchain based advanced data access control mechanism. The dynamic multi factor hashing scheme can change in response to the current Vulnerability of data and access patterns, whereas data access control refers to leverage blockchain's immutability and decentralized structure that helps protecting patient privacy while allowing authorized persons to read. The dynamic hashing method prevents intruder attempts by making hash and easy to calculate but requiring real-time modification of the hash for access protection. This is made possible by harnessing the application of zero-knowledge proofs (ZKP) within the frame of blockchain to enable verification of information when there is no disclosure of the data. Compared with the conventional methods, testing of prototype in a health care organization resulted in 92% on attempts by unauthorized workers to enter the system and 7% increasing data retrieval rate. These findings shows that the proposed model is a perfect patient data protection pattern in ehealth systems, because it is not only secures patients data but also enhances the accessibility and scalability to handle more clients. It is enabled by the use of zero-knowledge proofs (ZKP) in combination with blockchain technology to verify information, while keeping the information secret.
Keywords:	Decentralized, Block, Security, Access control, Attacks, Cybercriminals, ZKP, Multifactor
DOI:	https://doi.org/10.5281/zenodo.18259959
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	SWARMHEX-BCCLR: BIO-INSPIRED HEXAGONAL CLUSTERING FOR LONGEVITY OF IOT-WSN SURVIVAL TIME
Author:	KALAI KANNAN P , VIDHYA S
Abstract:	Wireless Sensor Networks (WSN)vis the base network for the advance IoT based network derivation. In WSNs, many sensor nodes work together to gather data from the environment. This data is first sent to a cluster head, which then passes it on to the base station for processing. In this paper, a nature-inspired clustering architecture known as SwarmHex, is proposed by integrating honeycomb and hexagonal structures to optimize Wireless Sensor Networks (WSNs). The proposed work enhances data transmission efficiency by structuring the communication pathway from individual nodes to cluster heads, and from cluster heads to the base station, thereby improving overall network longevity and energy utilization. The placement of nodes in the edges of the architecture provides the compact packing of sensors within a cluster with equal distance between the sensors tackle the coverage problem and transmission of data through the architecture using linear direction pattern with effective cluster head selection strategy based on priority helps the sensor nodes in efficient CH selection and data transfer. The proposed bio-inspired clustering SwarmHex-BCCLR shows best in class performance compared to the existing standard algorithms named as Ant Colony Optimization (ACO), Grey Wolf Optimization (GWO), Sandpiper Hybrid Optimization (SHO) and Whale Optimization Algorithm (WOA) that leads to longevity contribution to increased survival time of the WSN.
Keywords:	Bio-Inspired Clustering, Linear Pattern Routing, Survival Time, Base-Station, Honey Comb Hexagon Structure, Cluster-Head
DOI:	https://doi.org/10.5281/zenodo.18259977
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	LAGRANGIAN OPTIMIZATION-BASED DEEP Q-NETWORK FOR ROBUST AUTISM PREDICTION USING BEHAVIOURAL SCREENING DATA ACROSS AGE GROUPS
Author:	B. DEEPA , K.S. JEEN MARSELINE
Abstract:	Autism Spectrum Disorder (ASD) prediction remains a major challenge due to behavioural heterogeneity, data imbalance, and the absence of stable cues across age groups. Existing studies have relied on conventional classifiers or deep learning models that either fail to capture sequential dependencies in screening data or suffer from instability caused by noisy, incomplete, and imbalanced records. This gap highlights the lack of constraint-aware models that can both generalize and provide reliable decision confidence for early ASD detection. To address this, the present work introduces a Lagrangian optimization-based Deep Q-Network (LO-DQN), a novel reinforcement learning framework that embeds constraint-driven mechanisms entropy regularization, replay prioritization, Q-value bias suppression, and target synchronization. This design ensures stability in learning from diverse behavioural datasets and reduces the risks of overestimation and premature convergence common in existing methods. The study contributes new knowledge by demonstrating how constraint-guided reinforcement learning can be adapted to ASD screening contexts. Experimental results on structured, non-imaging datasets show improved classification performance and stability across varied diagnostic profiles. The findings establish LO-DQN as a scalable, interpretable, and constraint-resilient framework, opening a new pathway for early, low-cost, and clinically relevant ASD prediction.
Keywords:	Autism Spectrum Disorder, Screening Dataset, Deep Q-Network, Lagrangian Optimization, Behavioural Prediction, Reinforcement Learning
DOI:	https://doi.org/10.5281/zenodo.18259990
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	CT-LIVERNET: AN EFFICIENT DEEP LEARNING MODEL FOR LIVER LESION SEGMENTATION AND DIAGNOSIS
Author:	DR. SIREESHA VIKKURTY, Dr. ROHITH BALA JASWANTH B, Dr. B. VAMSY KRISHNA, MUDRAGADA SUNEETHA, MADHAN KUMAR JETTY, VEERAMOHANA RAO REDDY, SATHISH KUMAR SHANMUGAM, Dr. C P PAVAN KUMAR HOTA, N.MOUNIKA, Dr. SIVA KUMAR PATHURI
Abstract:	Medicine has undergone substantial changes because of big data analytics combined with deep learning technologies which enable doctors to predict diseases better while monitoring patients and diagnosing conditions as well as delivering better treatments. Liver disease stands as a worldwide significant health challenge attributed to its multiple complicated aspects and elevated death rates as well as its multiple pathologic expressions. A combination of wrong medical diagnosis and insufficient medical care decreases survival expectancy dramatically. Detecting liver disease presents a difficult issue because some damaged liver areas can look unblemished due to minimal lesion distribution which causes experts to make incorrect classifications. The assessment of liver conditions with precision at an early stage proves essential for shrinking treatment deficiencies along with enhanced medical results. The progress of deep learning and medical imaging allowed researchers to develop new detection procedures for liver disease analysis. The presented research develops a novel approach to liver disease recognition that adopts YOLO (You Only Look Once) as its feature extractor and pairs this with Random Forest (RF) and XGBoost classifiers for superior classification accuracy. The proposed diagnostic methodology reached a 95% accuracy level which proved superior to Support Vector Machine (SVM) and Logistic Regression and yielded optimal diagnosis and classification results.
Keywords:	Liver Disease, YOLO,RF, XGBoost, Logistic Regression, SVM.
DOI:	https://doi.org/10.5281/zenodo.18260014
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	POSE-GUIDED MULTI-PART ATTENTION FRAMEWORK FOR VIDEO-BASED PERSON REIDENTIFICATION
Author:	RANA. S.M. SAAD, MONA. M. MOUSSA, NEMAT S. ABDELKADER, HESHAM FAROUK
Abstract:	Person Re-Identification (ReID) is a fundamental task in intelligent video surveillance, aiming to recognize individuals across non-overlapping camera views. Despite significant progress, video-based ReID remains challenging due to variations in illumination, occlusion, and dynamic background clutter. To address these challenges, we present a novel ReID framework that jointly exploits spatial and temporal cues for robust person retrieval. The proposed approach integrates multi-level spatial representations by combining global, intermediate, and fine-grained features. Global and intermediate information is captured through self-attention mechanism, while fine-grained details are derived from pose-based attributes to enhance discriminative power. These spatial representations are then fused and processed by a temporal self-attention module, which models inter-frame dependencies to effectively capture motion dynamics. A retrieval stage subsequently leverages the joint spatial-temporal features to achieve accurate identification. Comprehensive experiments on the PRID2011 and iLIDS-VID benchmark datasets demonstrate the effectiveness of the proposed method. The framework achieves consistent improvements over baseline approaches, with gains of 2.3% and 3.4% in mean Average Precision (mAP) on PRID2011 and iLIDS-VID, respectively. These results highlight the robustness and discriminative capability of the proposed architecture for real-world video-based ReID applications.
Keywords:	Video Person Re-Identification; Person Re-Identification; Heatmaps; Pose; Pose Detection; Video Surveillance.
DOI:	https://doi.org/10.5281/zenodo.18260049
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	EXPLAINABLE HYBRID MACHINE LEARNING PARADIGM FOR TRUSTWORTHY ADVERSARIAL DEFENSE
Author:	Dr BJD KALYANI, Dr PANNALA KRISHNA MURTHY, Dr Y LAKSHMI PRASANNA, Dr M PADMAVATHI, Dr SARABU NEELIMA, NITHIN YADAV MASHAM
Abstract:	The accuracy of user-generated suggestions is critical, especially in decision-making processes where AI systems play an important role. Explainable AI (XAI) plays a vital role in decision making process as a mechanism for assessing and verifying the validity of these inputs, ensuring that AI models stay correct and reliable. This focus concentrate on how to apply XAI approaches to determine if users provide correct suggestions or attempt to deceive the system, hence improving the accuracy of AI-driven decision-making. This paper recommends a novel hybrid model that incorporates numerous sophisticated techniques, including Noisy Student self-training, Generative Adversarial Networks (GANs), and XAI. The Noisy Student technique enhances model resilience by self-training on noisy data, whereas GANs provide genuine adversarial instances for data augmentation. Ensemble learning and robust optimization techniques improve model resilience, guaranteeing that it can withstand and respond appropriately to hostile inputs. Generally, XAI techniques contribute transparency, allowing for the discovery and mitigation of misleading inputs by clarifying model predictions. This hybrid method not only improves the robustness and accuracy of AI models, but it also sets a new bar for openness in AI systems, addressing both technical and ethical concerns in modern AI applications.
Keywords:	Artificial Intelligence, Generative Adversarial Networks (GAN), Explainable AI (XAI), Local Interpretable Model-Agnostic Explanations (LIME), Shapley Additive Explanations (SHAP).
DOI:	https://doi.org/10.5281/zenodo.18260079
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	INTERPRETABLE DEEP LEARNING FOR IIOT: AN LSTM–IFWT DRIVEN SENSOR DATA ANALYTICS FRAMEWORK FOR E-WASTE MANAGEMENT AND SUSTAINABLE MONITORING
Author:	P. SANTHUJA, DR. A SURESH
Abstract:	Contextually, the Industrial Internet of Things (IIoT) contributes significantly to the Industry 4.0 to deliver intelligent monitoring and predictive maintenance systems. Nevertheless, the IoT sensor data, particularly in alpha-phase applications, are very volatile and intermittent with devices, which create significant obstacles to trustworthy data analysis. This paper introduces a mature and scalable framework to Long Short-Term Memory (LSTM) network and an innovative Intuitive Feature Wavelet Transformation (IFWT) mechanism that may be employed to successfully model and explain such time-varying time-series data. LSTM is a form of Recurrent Neural Network (RNN), and may be utilized to learn a temporal dependency between multivariate temperature measurements observed by sensors installed within as well as outside a controlled room atmosphere. The implementation using the architect of the Block chain layer achieved by the application of 5 layer and 13-layer functional transformations developed by using e-waste information security and threefold and multilabel categorization resolution in the advertisement of revolutionary IOT. Results of the experiment indicate that the suggested LSTM-based model is highly accurate in forecasting and detecting anomalies even in the presence of data instability. In order to make the models more interpretable, IFWT is proposed as a post-hoc approach to analysis that converts the weights of learned features into human-intuitive formats. The results of the Precision and performance in terms of F-1scores have been found to be more than 15% better in time series solution and more than 96% better in terms of classifications than SOA architectures. The proposed study proposes the vitality of IIOT ecosystems on block chain technology employed with augmented throughput imploring the many state-of-the-art methods and other deep learning structures to actualize the many solutions on IIOT.
Keywords:	Industrial Internet of Things (IIOT), IFWT (Intuitive Feature Wavelet Transform), Deep Learning (DL), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM).
DOI:	https://doi.org/10.5281/zenodo.18260123
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	INTEGRATING SEMI-SUPERVISED LEARNING AND TRANSFORMER MODELS FOR OFFENSIVE LANGUAGE DETECTION IN ENGLISH–TAMIL CODE-MIXED SOCIAL MEDIA
Author:	RAMA S , MYTHILI R
Abstract:	In todays world, social media has evolved far bigger than the press media, becoming the dominant medium for public expression. Despite the extraordinary growth, social media involves the prevalence of offensive language and harmful content. As there is no real content moderation or compliance checks, social media has become an open market for trading offenses. lack of stringent content moderation and compliance checks, social media has become a place where individuals are able to express things that are objectionable. In terms of translation requirements, particularly in low-resource code-mixed language pairs, such as English–Tamil, detection of offensive verbiage is quite a challenge due to factors, such as variations in scripts, irregular spellings, and non-availability of reliable large annotated corpora. This paper describes automated annotation pipeline for the detection of offensive language in English–Tamil translations using the XLM-RoBERTa-large transformer model. The goal of the proposed system is to generate high-quality multi-class offensive-language labels—without manual annotation. This model has been optimized for multi-class classifications in seven categories: profanity, hate speech, sarcasm, threats, harassment/bullying, disrespect/dismissive, and non-offensive. We hypothesize that a semi-supervised, confidence-tiered annotation process, combined with lexicon-based category mapping, public datasets that are labelled, machine-translated resources, and synthetic data generation, can achieve annotation quality comparable to human labelling. The imbalance in classes is addressed through back-translation, generation of synthetic data, and optimization of class-weighted loss. Experimental results indicate that the model can achieve a Macro-F1 score of 92.4% and a Cohen’s Kappa of 0.92, demonstrating near-human annotation quality.
Keywords:	Semi-Supervised Learning, Annotation, Offensive Language Labelling, Low Resource Language, Multi-Class Classification
DOI:	https://doi.org/10.5281/zenodo.18260204
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	AUTONOMOUS ROBOTICS IN HAZARDOUS ENVIRONMENTS DEVELOPING AI ALGORITHMS FOR SAFE AND EFFICIENT OPERATION
Author:	YANDRA SRINIVAS, NARASIMHA RAO THOTA, ANNAPURNA GUMMADI, N SRIHARI RAO, D VENKATA RAVI KUMAR, SREENIVASULU BOLLA, K.MAKANYADEVI
Abstract:	This paper describes a new hybrid system that utilizes AI to enhance the safety, efficiency, and adaptability of autonomous robots in hazardous situations. Robots can move autonomously in complex and dangerous areas, as the proposed method relies on RL, CNNs, and real-time hazard detection. In several segments, the system was tested through simulations conducted in regions with radiation, chemical waste, and hazardous soil. The research findings demonstrate that this model yields shorter task times and nearly half the number of safety issues compared to previous methods. Besides, the robot moved away from danger on its own when hazards appeared on its path. The research highlights that AI-powered autonomous systems enable safer and more efficient work in disaster response, as well as within industries and military operations. It leads to significant advantages for robots made to operate in dangerous situations.
Keywords:	Autonomous Robotics, Hazardous Environments, Reinforcement Learning, Hazard Detection, Path Planning, Safety Protocols
DOI:	https://doi.org/10.5281/zenodo.18260226
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	IUX -ADNET: MULTI-FUSION DATA ANALYSIS WITH ATTENTION APPROACH COMBINED WITH IUX DENSE MODEL FOR ACCURATE COTTON YIELD PREDICTION
Author:	PORANDLA SRINIVAS, DR. A SURESH
Abstract:	Optimized approaches in agricultural productivity are known and critical factors to predict and identify the precise cotton yield prediction. This paper incorporates an innovative model architecture with the IUX framework using custom functionalities (I (identify), U (SDE), and X (explicit-DE)) implemented to integrate the multi-fusion approach on data. The proposed methodology describes the three models (i) IUX-Dense Neural Net, (ii) IUX-Attention model, and (iii) IUX-Fused Model, which are concatenated with both IUX-Dense and IUX-Attention models. These models are extensively trained with a multimodal-fused dataset (Mendeley dataset + weather-parameter-based synthetic dataset) that is imbued with different features. The experimental evaluations are implemented with 25 epochs, and its performance stability with early stopping criteria demonstrates the IUX-fused approach is far superior with 99.99% accuracy with both disease classification and yield predictions with the IUX-optimized solution with the least MSE (0.05), RMSE (0.24), and MAE (0.188) and the optimal EVS (explained variance score) (97.58). Similar comparisons with machine learning and other SOA architectures are made and tabulated with performance results on both classification and yield prediction analysis. Finally, the proposed framework with the IUX model approach, as IUX-ADNET demonstrates, has far superior capabilities and offers scalable and more reliable solutions towards cotton yield predictions and disease classification.
Keywords:	Stochastic Differential Equations (SDE), Differential Equations (DE), Explained Variance Score (EVS), Identify Feature with Stochastic Explicit differential Equations (IUX), Deep Learning (DL), Machine Learning (ML)
DOI:	https://doi.org/10.5281/zenodo.18260349
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	A NOVEL DATA SECURE MODEL FOR INTERNET OF HEALTH THINGS WITH A NEW LIGHTWEIGHT CRYPTOGRAPHY ALGORITHM AND STEGANOGRAPHY TECHNIQUE
Author:	SANGEETHA SUPRIYA KOLA, DR.JENO LOVESUM S P
Abstract:	Ensuring the security of data in Internet of Things (IoT) based healthcare systems (HS) presents considerable challenges due to the limitations of traditional embedding methods and cryptography techniques, leading to more memory consumption, more execution time, less security, inadequate payload capacity, and performance inefficiencies. To address these issues, the Bernoulli Fish-based Stego Algorithm (BFBSA) is introduced as an innovative solution. Specifically designed for IoT healthcare data, this algorithm is validated through the encryption and embedding of healthcare data. The process involves initializing IoT healthcare data, encrypting it using the BFBSA algorithm, and embedding the encrypted data within steganographic images. Performance analysis is conducted using key metrics such as payload capacity, encryption time, memory usage, PSNR, and MSE. Comparative analysis with existing approaches highlights the BFBSA model’s efficiency and its effectiveness in ensuring secure and optimized data management in IoT healthcare environments.
Keywords:	IoT Healthcare, Encryption, Embedding Image, Data Security, Crypto Algorithm.
DOI:	https://doi.org/10.5281/zenodo.18260371
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	A LIGHT WEIGHT MULTI HEAD SELF ATTENTION MECHANISM TO CONVOLUTION NEURAL NETWORK TOWARDS DOMAIN INCREMENTAL CHEST X-RAY IMAGE CLASSIFICATION
Author:	KAVITHA S, DR.H.HANNAH INBARANI
Abstract:	In recent year, Chest X-ray images have been widely adopted towards pathological diagnosis of the respiratory diseases and pulmonary diseases. Advances in computer vision approaches and artificial intelligence offers promising solutions with high accuracy and low computation overhead to disease classification and severity estimation. Traditionally Chest X-ray images classification has been performed using machine learning and deep learning model architectures to provide excellent performances as it incorporates domain incremental learning. Despite of several advantages of those models, domain incremental learning faces catastrophic forgetting challenges due to large domain gaps. In this paper, light weight multihead self-attention mechanism which is considered as transformer architecture is incorporated into ResNet 50 based convolution neural network to fine tune strategies to learn and calculate the domain invariant features and domain specific features. Especially attention mechanism performs feature fusion among different domains to prevent forgetting. Further multi-label contrastive loss is built and trained as loss function to enhance class distinctiveness of the disease classes. Initially Convolution layer of the ResNet architecture is applied to extract the hierarchical feature from X-ray images. Next, Extracted features were projected to pooling layer incorporating the attention mechanism. Attention mechanism use attention coefficient to compute the domain invariant features and domain specific features to establish feature maps. Those feature map of different domains are aggregated and classified in the fully connected layer composed of activation function, normalization function, softmax function and loss function. Activation function using ReLu function aggregates the feature maps while softmax function using Naïve bayes classifier classify the feature map into disease categories as COVID-19 variants, Normal, Viral Pneumonia, and Lung Opacity. Experimental results and performance results substantiate the effectiveness of the current architecture with accuracy of 98.9% and surpassing the baseline methods exclusively in classifying the Chest x –ray images into multiple classes. Finally results emphasize the significance of the transformer architecture for optimizing the conventional deep learning architecture in the medical imaging.
Keywords:	Transformer, ResNet50, Multi-Head Self Attention, Convolution Neural Network, Chest X-ray Images, Respiratory Diseases, Pulmonary Diseases, Medical Imaging.
DOI:	https://doi.org/10.5281/zenodo.18260398
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	DEEP NEURAL NETWORK-BASED APPROACH FOR DIABETIC FOOT ULCER CLASSIFICATION
Author:	MOHAMMAD BANI YOUNES , MOHAMED S.SAWAH
Abstract:	Diabetic Foot Ulcer (DFU) is a severe complication of diabetes, necessitating early detection to prevent limb amputation. This study proposes a novel hybrid Convolutional Neural Network (CNN) model leveraging EfficientNetV2S architecture for automated DFU classification. Utilizing a dataset of 1,055 foot skin patches (512 abnormal, 543 normal), the model integrates transfer learning, dropout layers, global average pooling, and cosine decay learning rate optimization. Pre-trained weights from EfficientNetV2S were fine-tuned, achieving a state-of-the-art accuracy of 99.94% on the test set. Training and validation curves demonstrated stable convergence, with minimal overfitting, underscoring the model’s robustness. Comparative analysis with existing methods highlights superior performance, suggesting its potential for clinical deployment in early DFU diagnosis. This approach enhances diagnostic precision, enabling timely intervention to improve patient outcomes.
Keywords:	Diabetic Foot Ulcer; Deep Learning; EfficientNetV2S; Medical Image Classification; Transfer Learning.
DOI:	https://doi.org/10.5281/zenodo.18260414
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	CHATBOT INTEGRATION IN GAMIFIED E-LEARNING: BRIDGING ONE-SIZE-FITS-ALL AND ADAPTIVE APPROACHES
Author:	ZECRI EZZOUBAIR, EL HADDIOUI ISMAIL, OUZZIF MOHAMMED
Abstract:	It is true that human interaction can play a significant role in the learning process. The lack of these interactions in e-learning can lead to feelings of isolation and disengagement among learners. Several researchers have proposed gamified e-learning system model as an effective way to increase learners' motivation by providing them with autonomy, relatedness, and competence. Basing on literature, we can divide gamification models into three categories: - The "one size fits all" model where the game elements and mechanics are pre-determined and applied to all learners in a fixed manner. - The static adaptation models refer to a gamification approach where the game elements and mechanics are pre-determined and applied to all learners in a fixed manner, according to their profile. - The dynamic adaptation models refer to an approach of gamification where mechanics are adapted in real-time to the learners' changing needs. In order to help designers of gamified learning management systems to better maintain the learner’s motivation, we propose a new model for a gamified learning system that incorporates game elements using the three models explained above. This model aims to address the limitations of traditional e-learning platforms by providing a more engaging and personalized learning experience for students.
Keywords:	Adaptive Gamification, Learner engagement, E-Learning, Instructional Design, Adaptation Model
DOI:	https://doi.org/10.5281/zenodo.18260432
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	VISUAL DESCRIPTION GENERATING SYSTEM BY USING CNN AND BI-LSTM
Author:	ANIL KUMAR MUTHEVI , MANIKYALA RAO BOLLU , BHEEMA RAO RVVN, MAGANTI VENKATESH, NALLAMILLI V K REDDI, LANKOJI V SAMBASIVARAO
Abstract:	Image captioning plays a vital role for numerous purposes. The images contains the captions can lead to rapid and descriptively precise image searches and indexing. Unidirectional LSTM (Long Short-Term Memory) decoders, commonly used in existing systems, have limitations in retaining past information, leading to less than optimal results for long sequences In recent years, the advent of deep neural networks has made image captioning a viable concept. This Visual description generator system (VDGS) introduces a model grounded in deep learning to generate captions for input images. The model takes an image as input, employing algorithms such as Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (Bi-LSTM). The CNN component identifies objects within the image, while the Bi-LSTM constructs sentences about the input image, resulting in a caption tailored to the project. The proposed model primarily emphasizes object identification and formulating the most fitting title for input image. This study presents an enhanced Visual Description Generating System (VDGS) using CNN-based feature extraction and Bi‑LSTM‑based caption generation. The novelty lies in integrating bidirectional contextual modelling, improved sequence handling, and enriched semantic alignment between images and text. The system overcomes limitations of unidirectional LSTMs by incorporating forward–backward temporal processing, improving coherence for long descriptions .The novelty of this work lies in combining CNN (InceptionV3) feature extraction with a Bi‑LSTM decoder to capture both past and future contextual cues. This dual‑context modelling strengthens semantic coherence, improves caption fluency, and enhances alignment with visual regions.
Keywords:	Object Identification, Image Caption, Recurrent Neural Network, Convolution Neural Network Bi-directional Long-Short Term Memory
DOI:	https://doi.org/10.5281/zenodo.18260451
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	ACO–RNN: A MULTI-OBJECTIVE ANT COLONY OPTIMIZATION AND RECURRENT NEURAL NETWORK FRAMEWORK FOR DISASTER TWEET SUMMARIZATION
Author:	FAHD A. GHANEM, M.C. PADMA, MOHMMED A. S ALMOHAMADI
Abstract:	The rapid growth of disaster-related content on social media offers valuable opportunities for situational awareness, yet it also creates major challenges due to noise, redundancy, and the rapidly evolving nature of crisis events. Existing disaster tweet summarization approaches, including ontology-based, embedding-based, and neural methods, typically address semantic relevance, redundancy reduction, or coherence in isolation. As a result, they struggle to simultaneously control redundancy, preserve temporal coherence, and remain scalable for real-time disaster response.This study is motivated by the need for a unified summarization framework that jointly addresses these challenges within a single, scalable, and unsupervised model. To this end, we propose Ant Colony Optimization and Recurrent Neural Network (ACO-RNN), a hybrid framework that integrates Word2Vec embeddings for semantic representation, Ant Colony Optimization (ACO) for redundancy-aware sentence selection, and a Recurrent Neural Network (RNN) to preserve sequential coherence across selected tweets. The summarization task is formulated as a multi-objective optimization problem that balances relevance, coverage, and diversity while minimizing redundancy. The proposed framework is evaluated on four benchmark disaster tweet datasets, namely TyHagupit, SandyHShoot, HydBlast, and UFlood, using Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. Experimental results show that ACO-RNN achieves ROUGE-1 F1-scores ranging from 0.587 to 0.614, with relative improvements of 8 to 13 percent over strong baseline methods, including OntoDSumm and IKDSumm, along with consistent gains in ROUGE-2 and ROUGE-L. These findings demonstrate that combining redundancy-aware optimization with sequential neural modeling effectively overcomes key limitations of existing approaches and produces concise, coherent, and informative summaries suitable for disaster response and situational awareness.
Keywords:	Disaster Tweets; Word2Vec; Text Summarization; Ant Colony Optimization; Recurrent Neural Networks.
DOI:	https://doi.org/10.5281/zenodo.18260482
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	QUANTUM-ASSISTED HYBRID SECURITY FRAMEWORK FOR SECURE AUTHENTICATION WITH INTRUSION DETECTION FOR 5G NETWORKS
Author:	AANANDARAM V, P. DEEPALAKSHMI, VIDHYA PRAKASH RAJENDRAN
Abstract:	5G networks provides a lot of connectivity and are very advanced but also brings a lot of security challenges. These challenges include managing keys that can resist quantum attacks, making sure users are secure and detecting unwanted access in real time. Traditional security methods and machine learning tools are not safe against both regular and quantum-based attacks. This work introduces QSA-IDF (Quantum-Assisted Secure Authentication and Intrusion Detection Framework), a BB84-based Quantum Key Distribution hybrid security framework for quantum-driven secure authentication and Variational Quantum Classifier (VQC) for quantum-driven intrusion detection. The framework is simulated using IBM Qiskit and implemented on MEC-edge layers using NS-3 and Mininet, simulating actual 5G slices. Experimental findings on CICIDS 2018 dataset show 99% detection accuracy, 98% F1-score, and higher quantum fidelity scores than classical baselines. Comparison with SVM, RF, and CNN models indicates the scalability and robustness of the quantum system. The contribution of the paper lies in offering a fully integrated, end-to-end quantum-secure security framework solving key limitations of existing 5G security architectures and providing scalable efforts towards 6G-ready quantum-secure communications.
Keywords:	5G, Network Security, Quantum, Machine Learning, QKD, QML, Alice, Bob, Eavesdropping, Intrusion Detection.
DOI:	https://doi.org/10.5281/zenodo.18260517
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	AN NLP-BASED APPROACH FOR SARCASM DETECTION IN MARATHI LANGUAGE WITHIN MULTILINGUAL ENVIRONMENTS
Author:	PRAVIN PATIL, SATISH KOLHE
Abstract:	The expansion of social media platforms has required the development of complex natural language processing (NLP) methods for sentiment analysis and sarcasm detection, especially for low-resource languages. This research presents a novel, ensemble-based NLP framework for sarcasm detection in Marathi-English code-mixed text, addressing a significant gap in multilingual sentiment analysis. The study confronts the challenges inherent to the Marathi language, including its rich morphology and a paucity of annotated corpora, by constructing a dedicated dataset of 2,400 tweets. Through a hybrid annotation strategy, this corpus was refined to contain 647 sarcastic, 1,477 non-sarcastic, and 276 unsure instances. The primary aim of this study is to develop and evaluate an ensemble-based model that enhances sarcasm detection accuracy in Marathi-English code-mixed text through the integration of traditional machine learning and neural network approaches. The proposed methodology employs comprehensive feature engineering—incorporating TF-IDF, word embedding, sentiment lexicons, and code-switching indicators—and an ensemble model architected to synergize the strengths of a Multinomial Naïve Bayes classifier and a Spiking Neural Network (SNN). Empirical evaluation demonstrates that the ensemble model achieves superior performance, attaining an accuracy of 94.2%, a precision of 90.5%, a recall of 92.1%, and an F1-score of 91.3%. These results signify a substantial improvement over established baseline models, including XGBoost (87% accuracy), SVM (85% accuracy), BERT-base Multilingual (92% accuracy), and XLNet (90% accuracy). This work validates the efficacy of ensemble learning and context-sensitive feature extraction for sarcasm detection, providing a scalable and robust paradigm for analogous low-resource linguistic environments.
Keywords:	Sarcasm detection, Marathi NLP, Code-Mixed Text, Ensemble Learning, Multilingual Sentiment Analysis, Low-Resource Language
DOI:	https://doi.org/10.5281/zenodo.18260545
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	UNMASKING DEEPFAKE FACE SWAPS USING MULTI-RESOLUTION HYBRID DCT-DWT APPROACH
Author:	BHAVANI RANBIDA, DEBABALA SWAIN, RISHI RAJ
Abstract:	In todays digital world, the advanced tools for image editing using AI technologies have made it harder to identify whether an image is real or fake. Deepfake face swaps are the most convincing forms of image forgery among different manipulation formats. This study presents a new hybrid method that combines the Discrete Cosine Transform (DCT) and the Discrete Wavelet Transform (DWT) to detect such deepfakes. By breaking the image into different frequency and spatial layers, the proposed method can easily detect the visual inconsistencies happened due to the manipulations. By combining both DCT and DWT, the manipulated regions can be efficiently located. The experiment shows that the proposed hybrid model improves both accuracy and computational performance. It offers a robust and reliable framework for digital image authenticity against the growing deepfake technologies.
Keywords:	DCT, DWT, Digital Image Forensic, Deepfake, Multi-Resolution, Counterfeit Detection.
DOI:	https://doi.org/10.5281/zenodo.18260569
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text

Title:	EVALUATING THE CODE GENERATION CAPABILITIES OF CHATGPT THROUGH C++ DATABASE MANAGEMENT TASKS
Author:	MOHAMMED ALABDULLATIF
Abstract:	Large language models (LLMs) are becoming important tools in software engineering. These models, trained on large amounts of text, can understand and generate natural language with high accuracy. Today, they are being used across many stages of software development, from improving code quality and generating documentation to supporting natural language interaction with development tools. Among their many applications, code generation is one of the areas where these models have shown the greatest impact. This study explores the coding capabilities of ChatGPT GPT 4, a large language model introduced by OpenAI in May 2024. The model was evaluated on twenty C++ database management tasks, and the quality of its outputs was assessed using Cppcheck, a static code analysis tool. The results show that ChatGPT consistently produces executable C++ programs, though differences in code length and structure highlight its inherent nondeterministic nature. Most of the issues detected by static analysis were stylistic rather than functional or logical. Overall, this work provides empirical evidence on both the reliability and the current limitations of LLM based code generation, offering directions for future research.
Keywords:	Software Engineering, ChatGPT, Large Language Models, Coding, Programming Languages
DOI:	https://doi.org/10.5281/zenodo.18260582
Source:	Journal of Theoretical and Applied Information Technology 15^th January 2026 -- Vol. 104. No. 1-- 2026
Full Text