International Journal on Recent and Innovation Trends in Computing and Communication

A Study on Efficient Ai-Enabled Virtual Classroom

T. Parimalam — 2026-09-10

The global shift toward online and hybrid education following the COVID-19 pandemic has redefined the learning landscape. In India, the National Education Policy (NEP) 2020 champions the democratization of technology-enabled education. This vision aims to build an inclusive digital infrastructure across all higher education institutions nationwide. Despite these mandates, a significant digital divide remains a persistent reality for rural autonomous colleges. Statistics reveal that roughly 64% of these institutions struggle with systemic barriers to technological adoption. One major hurdle is the bandwidth deficit, where heavy video tools collapse under erratic 2G or 3G connectivity. High-definition streaming remains out of reach for many students in regions with limited internet infrastructure. Linguistic disenfranchisement also poses a psychological barrier for students from vernacular-medium backgrounds. English-centric interfaces create friction, preventing students from expressing complex ideas in their native tongues. Data privacy vulnerabilities further complicate the landscape as centralized servers process sensitive student information. Compliance with the Digital Personal Data Protection Act makes invasive video-based tracking a legal and ethical risk. Current academic literature and commercial products often fail to address these three critical dimensions simultaneously.

Predicting Student Academic Performance Using Combined Machine Learning Algorithms

P. Ramya — 2026-09-10

Accurate prediction of student academic outcomes supports early intervention strategies and enhances institutional planning. This study evaluates multiple machine learning algorithms—Random Forest, XGBoost, Support Vector Regression, Linear Regression, and Ridge Regression—using a comprehensive student dataset containing demographic, academic, behavioural, and lifestyle attributes. An ensemble voting regressor combining the five models shows improved accuracy over individual algorithms. Two sets of results are retained: the first demonstrating ensemble superiority with an R² of 0.32, and the second presenting regression coefficients and an alternative R² of 0.19. Across models, attendance, sleep patterns-habits, Behaviour, Regularity and previous grades emerge as significant predictors. The study highlights the importance of integrated modelling approaches and identifies key factors influencing academic performance.

An AI-Enabled Multi-Agent Reinforcement Learning Framework for Secure and Adaptive Routing in Vehicular Ad Hoc Networks

Infant Jansi I — 2026-09-10

Vehicular Ad Hoc Networks (VANETs) are an essential component of intelligent transportation systems, enabling communication among vehicles and roadside infrastructure for traffic safety and efficient transportation management. However, the highly dynamic topology, rapid mobility, and growing cybersecurity threats make secure and reliable routing a major challenge. Traditional routing protocols fail to adapt efficiently to changing network conditions and malicious attacks. To address these issues, this paper proposes an Artificial Intelligence (AI)-enabled secure adaptive routing framework using Multi-Agent Reinforcement Learning (MARL). In the proposed model, each vehicle acts as an intelligent autonomous agent capable of learning optimal routing decisions through continuous interaction with the environment. The framework integrates AI-based trust evaluation and anomaly detection mechanisms to identify malicious nodes and improve routing security. Reinforcement learning techniques optimize routing performance using parameters such as packet delivery ratio, delay, throughput, link stability, and trust values. Simulation results demonstrate that the proposed framework significantly improves packet delivery ratio, reduces end-to-end delay, increases throughput, and achieves high malicious node detection accuracy compared to conventional routing protocols. The study highlights the effectiveness of integrating AI and cybersecurity techniques for next-generation intelligent vehicular communication systems.

Next-Generation Data Mining Techniques for Healthcare, IoT, and Cybersecurity: A Unified Framework

Kiruthika P — 2026-09-10

The convergence of the Internet of Medical Things (IoMT), cloud computing, and pervasive healthcare systems has generated unparalleled volumes of real-time telemetry. However, this hyper-connectivity introduces critical vulnerabilities, making health networks premier targets for sophisticated cyber threats. Traditional data mining frameworks fail to balance real-time diagnostic mining with concurrent cryptographic security and threat detection. This paper proposes a unified, next-generation data mining framework that simultaneously extracts predictive clinical insights and neutralizes cybersecurity threats at the IoT edge. By integrating Federated Learning (FL), Generative Adversarial Networks (GANs) optimized via Adaptive Moment Estimation, and Homomorphic Encryption with Laplacian Differential Privacy, our system achieves high-accuracy diagnostic anomaly detection while maintaining a Zero-Trust cryptographic posture. Experimental simulations demonstrate a clinical pattern-extraction accuracy of 98.5% while keeping latency under critical thresholds for real-time IoMT deployment.

Hybrid Machine Learning-Based Data Mining Framework for Early Fraud Detection in Digital Transactions

R. Aravind — 2026-09-10

The exponential growth of global electronic payment systems has simultaneously catalysed a dramatic surge in sophisticated digital transaction fraud. Traditional rule-based engines and standalone machine learning classifiers face critical bottlenecks: they either struggle with high false-alarm rates or lack real-time computational scalability when handling massively imbalanced streaming datasets. This paper presents a novel, two-stage HybridMachine Learning-Based Data Mining Framework designed specifically for the ultra-early isolation of fraudulent patterns. The first stage employs unsupervised Autoencoders to conduct low-latency feature extraction and isolate non-linear transaction anomalies from heavily skewed datasets. The second stage feeds these optimal feature mappings into an optimized ensemble pipeline featuring Light Gradient Boosting Machine (LightGBM) to execute precise fraud classification. Tested against the standard benchmark PaySim mobiletransaction dataset, our hybrid architecture achieves a fraud classification accuracy of 98.4% and an overall recall of 97.8%, while cutting transaction assessment latency to just 2.6milliseconds. These outcomes systematically outperform legacy standalone classifiers, presenting an adaptive, risk-aware security layer ideal for high-throughput digital banking ecosystems.

Energy-Efficient Cross-Layer Optimization Framework for Wireless Sensor Networks Using Adaptive Routing and MAC Coordination

R.Janaranjani — 2026-09-10

Wireless Sensor Networks (WSNs) are distributed communication networks composed of small sensor nodes capable of sensing, processing, and transmitting data wirelessly. These networks are widely used in applications such as environmental monitoring, military surveillance, disaster management, healthcare systems, industrial automation, smart cities, and agricultural monitoring. Each sensor node generally consists of sensing components, microcontrollers, transceivers, and limited power sources. Due to their compact size and low deployment cost, WSNs have gained immense importance in both academic research and industrial applications.

A Novel Intelligent Tracking Framework for Wireless Sensor Networks based on Sammon Regularization and Circumference-Reinforced Projection Pursuit

S. Vasanth Kumar — 2026-09-10

In recent years, Wireless Sensor Networks (WSNs) used in extensive application ranging between military, surveillance, smart cities and so on. Target object tracking is one of the most fascinating applications in this domain of interest that specifically comprises of detecting the target object and keeps tracking of their movements. Several methods have been developed for target object tracking in WSN with minimal energy consumption. However, the accuracy level was not increased by existing tracking techniques. In order to address these problems, a novel targets object tracking method called Gaussian Distributive Sammon Regularization Policy (GG-DSRP) for efficient target object tracking in WSN is proposed. To make tracking more accurate another novel is also proposed called Circumference Reinforced Projection Pursuit and Adaptive Boosting (CRPP-AB) for tracking target object in WSN. The GG-DSRP method consists of three major processes namely reference node selection, target detection, and trajectory prediction. For this process Wilcoxon rank-sum test for identifying the reference node based on higher residual energy is applied in the first stage and in final stage, the target object trajectories are identified using Sammon projective on-policy learning algorithm to predict the target trajectories based on the state transition property. In CRPP-AB, node selection is done with Circumference Reinforced Acceleration and Adaptive Boost Target Object Classification is carried out to identify the target trajectory in WSN with Soft-Margin Support Vector Machine as the weak learners. Experimental evaluation is carried out on factors such as energy consumption and target object tracking accuracy with respect to different number of sensor nodes and data packets.

Artificial Intelligence in Urban Climate Prediction and Sustainability: A Systematic Review of Methods, Data, and Challenges.

Sivaranjani V — 2026-09-10

Urban climate variability has become a critical concern due to rapid urbanization and its impact on environmental sustainability. Artificial Intelligence (AI) has emerged as a powerful tool for modeling and predicting complex urban climate patterns by leveraging large-scale heterogeneous data. This paper presents a systematic review of recent advancements in AI-driven urban climate analytics, focusing on machine learning and deep learning techniques applied to temperature prediction, air quality assessment, and microclimate modeling.

A structured review methodology was adopted to analyze studies published between 2018 and 2025 from major scientific databases, including IEEE, Springer, Elsevier, and Scopus. The selected studies are categorized based on modeling approaches, data sources, and application domains. The review highlights the increasing adoption of convolutional neural networks (CNNs) and hybrid models for capturing spatial–temporal dependencies.

Despite notable progress, challenges such as limited data integration, poor model generalization, and lack of interpretability remain. This paper identifies research gaps and proposes future directions including multi-modal data fusion and explainable AI for sustainable urban climate management.

Lifecycle-Aware Post-Quantum Cryptography with Provenance Verification for Secure Cloud Microservices

S. Nithya — 2026-09-10

The transition from classical cryptographic schemes to post-quantum cryptography (PQC) in cloud microservices introduces significant challenges in maintaining performance, interoperability, and verifiable trust. This work presents Lifecycle-Aware Post-Quantum Cryptography with Provenance Verification for Secure Cloud Microservices (LAPQC), a novel framework that enables controlled and traceable cryptographic evolution across distributed service environments. Unlike existing approaches that rely on static replacement of cryptographic primitives, the proposed model introduces a lifecycle-aware transition mechanism combined with provenance-driven verification. The framework comprises two core modules. The Lifecycle-Aware PQC Transition Engine (LAPTE) facilitates adaptive switching between classical, hybrid, and PQC schemes based on operational context, ensuring seamless migration without service disruption. The Cryptographic Provenance and Verification Engine (CPVE) maintains verifiable records of cryptographic operations, enabling integrity validation, auditability, and protection against downgrade and substitution attacks. Experimental evaluation is conducted using metrics such as throughput and end-to-end latency under real-time workloads. Observed results indicate that while classical AES achieves maximum throughput, the proposed hybrid PQC scheme maintains competitive throughput levels while significantly improving quantum resilience. Furthermore, the hybrid approach demonstrates a balanced trade-off between performance and security compared to standalone lattice-based and hash-based PQC methods, which exhibit higher latency overheads. The integration of lifecycle-aware transition with cryptographic provenance introduces a new paradigm for secure and verifiable PQC adoption, making LAPQC a scalable and robust solution for next-generation cloud microservice architectures.

Automated Glaucoma Diagnosis Using Vision Transformer and Deep Learning Techniques on Retinal Fundus Images

S. Sangeetha — 2026-09-10

Glaucoma is one of the leading causes of irreversible blindness worldwide. Early diagnosis is critical to preventing permanent optic nerve damage. Traditional glaucoma diagnosis relies heavily on ophthalmologists’ expertise and manual examination of retinal fundus images, which can be time-consuming and subjective. This paper proposes a Novel Glaucoma Detection framework using Advanced Deep Learning and Vision Transformer (ViT)-based architectures for automated analysis of ophthalmic fundus images. The proposed system integrates image preprocessing, optic disc localization, data augmentation, and a hybrid Swin Transformer–CNN classification model for accurate glaucoma screening. Publicly available datasets such as REFUGE, ORIGA, and Drishti-GS are utilized for evaluation. Experimental results demonstrate superior performance compared with conventional CNN models. The proposed framework shows strong generalization capability and can assist ophthalmologists in early glaucoma diagnosis and large-scale screening programs.

IoT-Based Railway Track Fault Detection and Localization Using Acoustic Analysis

Sandra Jose — 2026-09-10

Railway transportation is one of the most efficient and widely used modes of transport worldwide; however, track faults remain a major cause of derailments and accidents. Traditional inspection techniques are manual, time-consuming, and prone to human error. This paper proposes an Internet of Things (IoT)-based system for railway track fault detection and localization using acoustic signal analysis. The proposed system employs distributed acoustic sensors integrated with microcontrollers to collect real-time sound signals from railway tracks. Machine learning techniques are utilized to classify faults based on extracted acoustic features. Experimental analysis demonstrates that the proposed system achieves high accuracy in identifying various track defects and enables real-time localization, thereby improving railway safety and maintenance efficiency.