Journal of Online Engineering Education

Improved Identification of Polycystic Ovary Syndrome by Ultrasound Imaging Employing Multi-Stage Image Processing and Deep Learning-Driven Segmentation with Hybrid Classification Methods

2026-08-17T11:05:41+00:00

Polycystic Ovary Syndrome (PCOS) is a widespread hormone illness impacting women of adulthood, usually defined by the emergence of many follicles within the ovaries and hormone dysregulation. Timely and precise identification by ultrasound imaging is essential to successful assessment & therapy. This research introduces a computerized method for the diagnosis & categorization of PCOS utilizing images and ML methodologies. The proposed framework starts with the preparation of ovary ultrasound images utilizing noise removal techniques, including Speckle Reducing Anisotropic Diffusion (SRAD), and contrast enhancement by Contrast Limited Adaptive Histogram Equalization (CLAHE). The improved pictures undergo U-Net segmentation technique to figure out the area of interest (ROI) and detect follicular features. Following that, mobileNetV2 technique is used for feature extraction. MobileNetV2 recovers organizational characteristics at many levels, encompassing low-level boundary details, mid-level texture patterns, and high-level semantic illustrations of ovary architecture. Classification is ultimately performed using machine learning models, such Support Vector Machine (SVM) and Convolutional Neural Network (CNN) to differentiate between PCOS and normal instances. Experimental findings indicate that the suggested design enhances diagnostic precision.

Intelligent Fraud Detection in Financial Transactions using Machine Learning Models

2026-08-17T11:12:20+00:00

The financial industry has a lot going on online, which brings up several problems. Fraudulent financial transactions are among the biggest concerns for individuals and financial institutions. Detecting fraudulent activity is crucial for banks and individuals. Traditional fraud detection systems do not effectively work with this large amount of transactional data, necessitating efficient fraud detection mechanisms for individuals and financial institutions worldwide. The financial industry has significantly reduced fraud due to the rise of technology and economic growth. This study shows that machine learning techniques have promising results in detecting fraudulent transactions. This paper proposes Parameterized Random Forest and other classifiers for fraud detection in financial transactions. We developed the Synthetic Minority Oversampling Technique (SMOTE)-Oversampling based machine learning model for monitoring financial transactions and detecting fraudulent activity. The proposed model uses Principal Component Analysis(PCA) to extract features from the transaction data and make predictions about the probability of fraud. We compare the suggested model's performance with other cutting-edge machine learning methods and assess its effectiveness on a dataset that is accessible to the general public. Our results show that the proposed model outperforms existing methods in accuracy and F1-score.

A Structured Comparative Analysis of Cloud, Fog, and IoT Computing with Performance Metric Classification

2026-08-17T11:15:13+00:00

Rapid increase in Internet of Things (IoT) devices has created a need for fast and efficient data processing. Cloud computing has been widely used because high computation and storage are provided, but delay is introduced due to centralized data handling. Fog computing has been proposed to reduce this delay by moving processing closer to data sources.
In this paper, a comparative study of Cloud, Fog, and Internet of Things (IoT) systems is presented. Differences in architecture, processing capability, latency, and resource management are examined in a clear manner. A layered structure is described to show how IoT devices, fog nodes, and cloud systems work together for data processing. Important performance metrics such as latency, energy consumption, resource utilization, and Quality of Service are also discussed for system evaluation.
It is observed that IoT devices mainly generate real-time data with limited resources, cloud systems handle large-scale processing with higher delay, and fog computing provides intermediate processing with reduced latency. Integration of these paradigms is necessary to support real-time and scalable applications such as smart cities, healthcare monitoring, and industrial systems.

A Unified Zero-Trust Framework Integrating Edge Intelligence, Blockchain, and Post-Quantum Cryptography for Secure 6G Cyber-Physical Systems

2026-08-17T11:17:46+00:00

The shift toward 6G networks brings advanced capabilities to Cyber-Physical Systems (CPS) and Internet of Everything (IoE) environments, but also introduces serious security concerns due to their highly distributed and dynamic characteristics. Existing perimeter-based security approaches are no longer sufficient to handle these challenges.
This study introduces a unified Zero-Trust security framework that combines Edge Intelligence, Blockchain, and Post-Quantum Cryptography (PQC) to achieve secure, efficient, and scalable communication. AI-driven Edge Intelligence, utilizing Deep Learning-based anomaly detection and adaptive AI techniques, enables rapid anomaly detection and adaptive threat mitigation at the edge. Blockchain technology supports decentralized authentication, ensures tamper-resistant record-keeping, and enables secure data exchange. In addition, PQC techniques, particularly lattice-based cryptography, are employed to protect systems from potential quantum computing threats.
The proposed framework is assessed using key metrics such as latency, communication overhead, detection performance, and resistance to cyberattacks including Distributed Denial of Service (DDoS) and Man-in-the-Middle (MITM). The results demonstrate higher detection accuracy, lower attack success rates, and improved data integrity compared to traditional methods.
Overall, this work offers a robust and future-ready security solution, supporting the safe deployment of emerging 6G applications such as smart grids, autonomous systems, and digital twin technologies.

Survey on Self-Sustaining IoT Systems via Multi-Source Energy Harvesting and Adaptive Power Intelligence for Predictive Maintenance

2026-08-17T18:01:45+00:00

The rapid expansion of Internet of Things (IoT) technologies in industrial environments has created significant demand for energy-efficient and autonomous sensing systems capable of operating under resource-constrained conditions. Traditional battery-powered IoT devices face critical limitations, including limited operational lifetime, frequent maintenance, and increased deployment costs, particularly in remote and inaccessible industrial locations. To address these challenges, recent research has focused on integrating energy harvesting techniques, intelligent power management, and lightweight edge intelligence into Industrial IoT (IIoT) frameworks.
This survey paper presents a comprehensive review of self-sustaining IoT systems designed for predictive maintenance applications. The study examines various energy harvesting approaches, including solar, vibration,
thermal, and hybrid energy harvesting mechanisms, highlighting their efficiency, reliability, and suitability for industrial environments. In addition, the survey analyzes adaptive power-aware strategies that dynamically regulate sensing, computation, and communication operations based on real-time energy availability.
The paper further investigates the role of edge computing and lightweight machine learning models in enabling real-time anomaly detection while minimizing energy consumption. Existing intelligent energy management techniques, predictive energy allocation models, and low-power communication protocols are comparatively evaluated in terms of energy efficiency, computational overhead, response latency, and maintenance requirements.

Algorithmic Approaches to Enhancing Economic Intelligence through Data Mining

2026-08-17T18:04:44+00:00

The accelerating complexity of global financial systems demands analytical tools capable of distilling actionable insights from vast, heterogeneous economic datasets. This paper presents the Adaptive Algorithmic Economic Intelligence (AAEI) framework, a novel multi-layer data mining architecture that integrates temporal pattern mining, ensemble classification, and an economic signal fusion mechanism to enhance economic intelligence extraction. The proposed system introduces two domain-specific evaluation metrics: the Economic Intelligence Quotient (EIQ) and Temporal Predictive Efficiency (TPE), which together assess the quality and timeliness of mined economic knowledge. Experiments conducted on the Federal Reserve Economic Data Monthly Database (FRED-MD) and the World Bank Economic Complexity Index (WorldBank ECI) demonstrate that AAEI achieves accuracy scores of 91.2% and 89.7%, respectively, outperforming established baselines including XGBoost, LSTM, SVM, and Random Forest by margins of 3.2% to 14.1%. The framework offers a principled pathway for deploying intelligent data mining in macroeconomic forecasting and financial decision support.

An Efficient Dual Attention Graph-based Recurrent Optimization Model for Software Defect Prediction

2026-08-17T18:06:41+00:00

Software defect prediction plays a critical role in improving software quality by identifying faulty modules early in the development process. This study aims to address the limitations of existing machine learning and deep learning models, such as their inability to effectively capture structural dependencies, temporal evolution, and high-dimensional feature interactions. To achieve this, a novel model named Dual Attention Graph-based Recurrent Optimization (DAGr-RO) is proposed. The model integrates graph-based learning to represent inter-module relationships, dual attention mechanisms to emphasize important features and structural connections, and recurrent neural networks (LSTM/GRU) to capture temporal patterns across software versions. Additionally, an optimization strategy is employed to dynamically update model parameters and enhance prediction performance. The proposed approach is evaluated using a Kaggle software defect dataset and compared with baseline models including RNN, CNN, Random Forest, and XGBoost. Experimental results demonstrate that DAGr-RO achieves superior performance with an accuracy of 94.5%, precision of 93.2%, recall of 92.6%, and F1-score of 92.9%, outperforming all comparative models. These findings confirm that the integration of graph learning, attention mechanisms, and temporal modeling significantly improves defect prediction. The study concludes that DAGr-RO provides an effective, scalable, and reliable solution for real-world software defect prediction tasks.

Machine Learning, Deep Learning Assisted Clinical Data Analysis on Myocardial Infarction

2026-08-17T18:08:09+00:00

Myocardial Infarction remains one of the leading causes of mortality worldwide, demanding accurate and timely diagnosis for effective clinical intervention. Traditional diagnostic approaches often rely on physician expertise, electrocardiogram interpretation, biomarker evaluation, and imaging results, which may delay rapid decision-making in emergency settings. Recent advancements in Deep Learning have demonstrated significant potential in transforming disease diagnosis through automated and data-driven analysis. This study presents a Deep Learning-assisted, clinical data-driven framework for the early detection and risk assessment of myocardial infarction using patient clinical records, laboratory parameters, and physiological indicators. The framework integrates data preprocessing, feature normalization, missing value imputation, and predictive modeling to improve diagnostic reliability. Several state-of-the-art Deep Learning models are evaluated for classification performance. These models are capable of learning complex nonlinear relationships among heterogeneous clinical variables and identifying hidden patterns associated with the occurrence of myocardial infarction. This paper discusses recent 10 years existing Machine Learning, Deep Learning methods and thereby guides further enhancements and improvements to state-of-the-art techniques. This research highlights the importance of intelligent healthcare analytics and demonstrates the Machine Learning and Deep Learning-assisted clinical decision systems in myocardial infarction diagnosis.

An Energy Centric Hierarchical Opportunistic Cognitive Routing Protocol for Reliable and Energy Efficient Wireless Sensor Networks

2026-08-17T18:09:56+00:00

Wireless Sensor Networks (WSNs) play a vital role in Internet of Things (IoT) applications such as smart cities, healthcare, agriculture, and industrial monitoring, where energy efficiency and reliable communication are critical challenges due to limited sensor node resources. Existing routing and clustering methods often suffer from high energy consumption, routing overhead, poor adaptability, and reduced scalability under dynamic network conditions. To address these issues, this paper proposes an Energy-Centric Hierarchical Opportunistic Routing Cognitive Protocol (ECHO-RC) for energy-efficient and reliable communication in WSNs. The main objective of the proposed framework is to improve network lifetime, packet delivery reliability, throughput, and spectrum utilization while minimizing delay and energy consumption. ECHO-RC integrates hierarchical clustering, opportunistic routing, cognitive spectrum awareness and utility-based adaptive decision making using residual energy, link quality and channel availability metrics for optimal cluster head and forwarding node selection. Simulation results demonstrate that the proposed ECHO-RC protocol achieves superior performance with a network lifetime of 1250 rounds, low energy consumption of 0.42 J, high packet delivery ratio of 96%, throughput of 310 kbps, and reduced delay of 118 ms compared with existing methods. The results confirm that ECHO-RC effectively enhances communication reliability, energy optimization, and overall network efficiency in dynamic WSN and IoT environments.

Efficient Computer Vision for Edge AI Applies Quantization Strategies that Optimize Memory, Latency, and Accuracy while Preserving Defect Sensitivity in Industrial Anomaly Detection under Limited Power Budgets

2026-08-17T18:11:49+00:00

Limitations on industrial anomaly detection in computer vision is moving to edge devices, where memory, compute throughput, latency, and power budgets often make full-precision deep models impractical. This paper reviews how edge constraints shape deployment choices and motivates model-compression pipelines that preserve defect sensitivity while enabling real-time inference. It surveys neural-network quantization as a core strategy for reducing model size and accelerating inference, covering post-training quantization and quantization-aware training, along with uniform and mixed-precision approaches and recent methods designed to minimize accuracy loss at low bit-widths. It then summarizes industrial anomaly-detection paradigms (unsupervised and supervised), common datasets and evaluation practices, and representative methods that rely on feature backbones and efficient scoring/localization. Finally, it connects algorithmic choices to practical deployment workflows on embedded platforms, highlighting toolchains and runtimes (e.g., INT8 execution paths) that translate quantized models into measurable gains in throughput and energy efficiency. A defect-sensitive 8-bit quantization perspective is discussed to illustrate how task-aware compression can retain anomaly-detection performance while meeting strict on-device constraints.

AI-Driven Human-Centric Decision Support for Resilient Predictive Maintenance in Asset Management Amidst Pandemics

2026-08-17T18:13:50+00:00

Pandemic-induced disruptions such as those experienced during COVID-19 have exposed critical vulnerabilities in conventional predictive maintenance (PdM) workflows, particularly within complex industrial asset management environments. Maintenance scheduling relies on technician availability, spare-part logistics, and sensor fidelity—all of which deteriorate sharply under large-scale health crises. This paper introduces an AI-Driven Human-Centric Decision Support System (AI-HC-DSS) that integrates a Transformer-LSTM hybrid deep learning architecture with Shapley Additive Explanations (SHAP)-based explainability and a Bayesian uncertainty quantifier to deliver transparent, actionable maintenance recommendations under pandemic-era constraints. The system embeds two novel evaluation metrics—the Pandemic Resilience Score (PRS) and the Mean Time-to-Failure Degradation Index (MTTFDI)—to capture operational continuity and failure-progression robustness across disrupted scenarios. Experiments conducted on the NASA C-MAPSS turbofan degradation dataset and the MIMII industrial sound anomaly dataset demonstrate that the proposed framework outperforms competing baselines, achieving an F1-Score of 0.931, AUC-ROC of 0.924, RMSE of 9.87 cycles, and a PRS of 0.930, representing improvements of 6.5%, 6.3%, and 47.4% over the next-best Transformer-only baseline, respectively. The results confirm that coupling AI predictive power with human-centric design and pandemic-aware modeling yields substantially more resilient maintenance operations.

Hierarchical Ensemble Learning with Adaptive Sampling for Rice Phenological Stage Detection under Class Distribution Skewness

2026-08-17T18:15:37+00:00

Accurate identification of paddy growth stages from satellite imagery is a prerequisite for precision agricultural management, yet the inherent temporal imbalance across phenological phases introduces severe classification bias in conventional machine learning pipelines. This study presents the Hierarchical Ensemble Learning with Adaptive Sampling for Rice Phenological Stage Detection (HELASRPD), a two-phase framework that couples cluster-based oversampling (CBO) with SMOTE-ENN hybrid resampling in the first phase, and a stacked hierarchical random forest ensemble with phenology-aware weighting in the second phase. Two publicly accessible benchmark datasets, the IRRI Paddy Dataset (IRRI-PD) and the NDVI Remote Sensing Paddy Dataset (NDVI-RS), are used for comprehensive evaluation. Two novel evaluation metrics, Phase-Weighted F-score (PWF) and Imbalance-Adjusted Kappa (IAK), are introduced to overcome the limitations of standard accuracy measures under skewed class distributions. The proposed model achieves 92.7% accuracy and a macro F1-score of 91.4% on IRRI-PD, outperforming all six competing baselines. Ablation experiments confirm that each architectural component contributes measurably to overall performance. HELASRPD demonstrates that systematic imbalance correction integrated within an ensemble hierarchy yields robust and interpretable phenological stage mapping.

Self-Supervised Federated Learning for Unlabelled Image Data with Contrastive Learning Mechanism an Overview

2026-08-17T18:18:23+00:00

The convergence of self-supervised learning, federated learning, and contrastive learning represents a paradigmatic shift toward privacy-preserving, decentralized machine learning systems that can effectively leverage unlabelled image data. This comprehensive review examines the theoretical foundations, methodological innovations, and practical implementations of self-supervised federated learning frameworks that incorporate contrastive learning mechanisms. The analyse the synergistic effects of combining these three learning paradigms, addressing critical challenges like data heterogeneity, privacy preservation, and model convergence through systematic analysis of recent developments identify key architectural patterns, algorithmic innovations, and empirical findings that demonstrate the potential of these integrated approaches. Our review reveals that contrastive learning mechanisms significantly enhance representation quality in federated settings while maintaining privacy constraints, though challenges remain in handling non-IID data distributions and communication bottlenecks. This work provides researchers and practitioners with a comprehensive understanding of current methodologies, identifies existing limitations, and proposes future research directions for advancing self-supervised federated learning systems.

Smart Hybrid Intelligent Encryption and Learning Defense an Adaptive Hybrid Ml-Cryptographic Framework

2026-08-17T18:20:35+00:00

The increase in the number and intricacy of cyberattacks is generating immensely difficult security environs for modern cryptographic systems that have already been challenged by cyberattacks, side-channel attacks, and now face the approaching threat of quantum computing. Traditional symmetric and asymmetric encryption algorithms can theoretically provide a high level of security from a mathematical standpoint, but they do not provide the necessary adaptive capabilities required to dynamically respond to the changing threat environment. In this paper, we propose a novel hybrid cryptographic framework called SHIELD (Smart Hybrid Intelligent Encryption and Learning Defense) that combines Machine Learning (ML) techniques with classical cryptographic primitives to implement adaptive and context-aware encryption. The SHIELD framework embodies a erratic Forest separator to classify real-time threats, a Convolutional Neural Network to detect a bizarre pattern in ciphertext traffic, and uses AES 256 and RSA 4096 as the foremost cryptographic engines. Additionally, the framework actively adjusts the key lengths, encryption modes, and algorithm option based on the recognized threat outline of the network environment. Experimental assessment on three datasets, NSL KDD, CICIDS 2017, and synthetic data construct from a custom generator, exemplify an overall accuracy rate of 98.7% for the SHIELD framework to classify threats, 23.4% lower than standard latency structure, and increased protest to known side-channel attacks. The framework was also significantly evaluated for multiple performance metrics and evaluated against other state-of-the-art techniques employ Precision, Recall, F1-Score, ROC-AUC, and computational overhead metrics.

An Edge-Enabled IoT Framework for Real-Time Stroke Risk Prediction Using Explainable Machine Learning

2026-08-17T18:22:17+00:00

One of the most common causes of death and long-term disability, stroke requires immediate detection. The current stroke prediction methods based on traditional machine learning methods are mostly based on offline datasets and the IoT-based healthcare systems mainly detect and monitor healthcare data without performing intelligent prediction. This study proposes a conceptual framework based on the Internet of Things (IoT) concept. Digital twins (IoT) and real-time health monitoring, edge-computing and explainable machine learning to dynamically forecast strokes. The suggested framework consists of a number of wearable sensors that continually monitor several physiological indicators such as heart rate, blood pressure, glucose level, and body mass index. The preliminary data preprocessing and anomaly filtering are done in an edge computing layer and the real-time stroke risk prediction is done by a cloud-based machine learning module. Especially, the framework includes explainable AI methods to enhance the understanding and trust in clinical decisions. The proposed system can offer “interpretable predictions” and low latency processing, and have continuous monitoring capabilities that make them suitable for next-generation smart healthcare applications when compared with conventional systems.

Advanced Medical Imaging using Quantum Sensing and Neuromorphic Processing with Space-Time Relativistic QKD based Secure Transmission

2026-09-03T09:53:45+00:00

Advanced medical imaging demands high sensitivity in signal acquisition, real-time intelligent processing and secure transmission of sensitive patient data. This paper presents a hybrid framework integrating quantum sensing, neuromorphic processing and space-time relativistic quantum key distribution [1] (STR-QKD) based secure transmission. Quantum sensing enables the detection of extremely weak biomedical signals beyond classical limits, enhancing imaging [2] precision. Neuromorphic processing, inspired by spiking neural networks, facilitates real-time and energy-efficient analysis of dynamic biomedical data. To ensure data integrity and confidentiality, STR-QKD [1] is used for secure key generation and transmission under relativistic constraints, enabling robust protection against eavesdropping. The proposed framework supports early detection of abnormalities, such as neurological disorders [8], while ensuring secure and reliable data flow across distributed medical systems. This integrated approach advances the development of intelligent, high-precision and secure next-generation medical imaging systems.

Yolo- Smart Fabriscan for Fabric Defect Detection for Improving Missed Detections of Visually Similar Defects

2026-09-03T09:55:56+00:00

Fabric defect detection remains a significant challenge, necessitating innovative approaches to enhance detection efficiency and accuracy. The accuracy of fabric detection is better achieved by YOLO based models compared to other Deep Learning (DL) models. However, some categories remain challenging, such as holes and drops defects. Due to the similarities in colour and texture features of these defects, the model tends to misclassify these defects. In addition to that combined with external environmental factors such as changes in lighting conditions, can increase the visual similarity between these two defects, thereby impacting the model’s classification accuracy. To improve the typical YOLOv8 models for defect detection classification, customized SPP variations and attention is proposed in this paper to differentiate between fine-grained characteristics. Spatial Pyramid Pooling with Multi-scale Attention and Local Aggregation Network (SPP-MALAN) is proposed. Multi-scale Attention in SPP can improve automatically adjusts the weight distribution in the feature maps to focus more on key areas of the image. This is particularly effective in accurately identifying similar defects in various scales under complex environments like changes in lighting conditions. Local Aggregation Network improves the model’s representational capacity by efficiently integrating features from different layers of SPP. Convolution, Batch Normalization and SwiGLU (CBSG) replaces CBS to understand context and identify complex similarities between defects. It provides greater expressivity, leading to better accuracy and convergence of learning model. The proposed model is named as YOLO- Smart FabriScan. Experimental results demonstrate that the proposed YOLO-Smart FabriScan model achieves superior accuracy of 97.01%, 97.06%, and 98.1% on the TILDA 400, AITEX, and Fabric Stain datasets, outperforming all compared models.