Volume 17, Issue 2, April 2026

AI-Driven Task Scheduling Across Edge–Fog–Cloud Architectures: A Comparative Review and Benchmark-Driven Design Blueprint

Pages: 1-6 (6) | [Full Text] PDF (216K)

Jawad Shafique

Department of Computer Science, University of Engineering and Technology Lahore, Pakistan

Abstract - Task scheduling in edge–fog–cloud (EFC) Internet of Things ecosystems is challenging due to heterogeneous resources, dynamic workloads, and competing objectives such as latency, energy consumption, reliability, and cost. Existing approaches range from analytical and optimization-based models to heuristic, metaheuristic, and AI-driven adaptive methods, but differences in system assumptions, workload models, objective functions, and evaluation practices make cross-study comparison difficult and often obscure robustness and security-related considerations. This paper presents a focused comparative review of thirteen peerreviewed studies on task and adjacent resource scheduling in EFC environments. To support structured and transparent comparison, we introduce EFC-SchedBench, a benchmark abstraction that harmonizes reported metric categories across performance, energy, cost, reliability, and security-related signals when available. Using this benchmark lens, we synthesize trends and trade-offs across the reviewed methods and identify persistent evaluation gaps. Based on these insights, we derive SHIELD-Sched as a benchmark-driven hybrid scheduling blueprint that integrates queue-aware admission control, fuzzy candidate pruning, adaptive deep reinforcement learning, and bounded-time heuristic fallback. The paper is positioned as a comparative review and design framework rather than an empirically validated system, and it aims to support reproducible and security-aware scheduling research in modern IoT deployments.

Index Terms - Task Scheduling, Edge Computing, Fog Computing, Cloud Computing, Internet of Things, Benchmarking, Deep Reinforcement Learning, Fuzzy Logic, Metaheuristics and Security.

Cloud-Connected Secure Healthcare Monitoring Platform for Portable ECG Monitoring

Pages: 7-12 (6) | [Full Text] PDF (343K)

Daniel Haou, Yukke Emmay

Department of Data Science and Software Engineering, Korea

Abstract - The pilot implementation of the system demonstrates both its practicality and efficiency, highlighting its capability to transform conventional ECG monitoring practices and improve patient care outcomes. Furthermore, the system can accurately capture ECG signals, calculate the total number of heartbeats, and classify heart rhythms as normal or abnormal. Such information is highly valuable for physicians in assessing a patient’s cardiovascular condition and recommending appropriate medical treatment. By enabling remote monitoring, reducing healthcare costs, and supporting early detection of cardiac abnormalities, the proposed system has the potential to save both time and lives. This paper discusses the five major advantages of the proposed ECG monitoring framework and presents recommendations for future improvements and development.

Index Terms - Healthcare, Secure Cloud, Monitoring and ECG

Improved Diabetes Prediction Accuracy Using Hybrid Machine Learning Models

Pages: 13-21 (9) | [Full Text] PDF (538K)

Qasim Zafar, M. Junaid Arshad

Department of Data Science, University of Engineering & Technology (UET), Lahore, Punjab, Pakistan
Department of Computer Science, University of Engineering & Technology (UET), Lahore, Punjab, Pakistan

Abstract - Diabetes mellitus (DM) is a chronic metabolic disorder affecting millions worldwide, and early detection is crucial for preventing severe complications and improving patient outcomes. Machine learning-based prediction systems have shown promise in diabetes risk assessment, but their performance is often degraded by the presence of missing values, outliers, and inadequate preprocessing in medical datasets. This research proposes a novel hybrid stacking ensemble framework for accurate diabetes prediction that addresses these limitations through comprehensive preprocessing and advanced ensemble learning techniques. The methodology employs the PIMA Indian Diabetes dataset comprising 768 female patients (268 diabetic, 500 non-diabetic). A robust multi-stage preprocessing pipeline is implemented: (i) group median imputation for handling missing values, (ii) IQR-based outlier detection with median replacement, (iii) max-min normalization to scale features between 0 and 1, and (iv) interaction feature engineering to capture complex relationships between variables. For classification, a hybrid stacking ensemble model is developed with five diverse base classifiers Random Forest (RF), XGBoost, LightGBM, AdaBoost, and Decision Tree (DT) and Logistic Regression (LR) serving as the meta-learner. The model is evaluated using 10-fold stratified cross-validation to ensure robustness and generalizability. Experimental results demonstrate that the proposed hybrid stacking ensemble achieves exceptional performance with an accuracy of 94.74%, sensitivity of 91.67%, specificity of 96.15%, precision of 91.67%, AUC of 0.9833, and F1-score of 91.67%. These results significantly outperform existing approaches. The proposed hybrid stacking ensemble model offers a reliable and accurate tool for early diabetes risk stratification, with potential for integration into clinical decision support systems. Future work will focus on validating the model on diverse datasets, exploring additional feature selection techniques, and developing a user-friendly web application for real-world healthcare deployment.

Index Terms - Diabetes, Missing Values, Outliers, Normalization, Interaction Feature Engineering, Machine Learning and Feature Selection

Wireless Communication Technologies in Internet of Things (IoT): A Comprehensive Review

Pages: 22-26 (5) | [Full Text] PDF (316K)

Ali Ilyas, Ghulam Murtaza

Department of Computer Science, University Of Engineering and Technology, Lahore, Pakistan

Abstract - The Internet of Things (IoT) has become a major enabling paradigm to the next-generation smart systems due to its mass connectivity offering between heterogeneous devices via wireless communication technologies. IoT solutions based on wireless communication are sensitive to the performance, scalability, and reliability of the applications. The following paper provides a detailed and organized review of wireless communication technologies applied to the IoT, such as short-range protocols, low power wide areas networks (LPWAN), cellular based solutions (LTE-M, NB-IoT, and 5G networks). It gives a detailed analysis of internet of things architecture, protocol stacks, security and energy issues and deployment issues. Moreover, there are widespread comparative analysis with performance indicators, sphere of application, and economical aspects that are introduced with the help of tables and high-tech figures. Lastly, the paper presents the research challenges and directions in the future with regard to intelligent and sustainable systems of sixth-generation (6G)-enabled IoT communication systems.

Index Terms - Internet of Things, Wireless Communication Technologies, LPWAN, 5G, 6G and Challenges

An Explainable Fraud Detection Technique for Credit Card Transactions using Machine Learning

Pages: 27-34 (8) | [Full Text] PDF (660K)

Tehreem Zahid

Department of Computer Science, University of Engineering & Technology, Lahore

Abstract - The high volume of electronic transactions in the contemporary financial system has heightened fraudulent practices, particularly in credit card transactions. In our previously published study, fraudulent transactions were detected using various machine learning algorithms, with good performance achieved by Logistic Regression and Bagging. Deep learning models like the Artificial Neural Network (ANN), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN) are applied in this extended research, as well as ensemble methods to enhance detection. Moreover, Explainable Artificial Intelligence (XAI) methods are used to perform understandable and transparent predictions of the model. The study aims to compare machine learning, ensemble learning, and deep learning models of credit card fraud detection as well as enhance interpretability. It has been found that deep learning algorithms, in particular CNN, are more accurate and have better explainability.

Index Terms - Machine Learning, Deep Learning, Credit Card Fraud Detection, Ensemble Methods, Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and Explainable Artificial Intelligence (XAI)

Integration of IoT with Artificial Intelligence

Pages: 35-40 (6) | [Full Text] PDF (411K)

Zaviyar Hasnain Bhutta

Department of Computer Science, UET, Lahore

Abstract - One of the most transformative trends in contemporary technology is the combination of Artificial Intelligence (AI) and the Internet of Things (IoT). The Internet of Things (IoT) enables billions of physical devices worldwide to collect and exchange data, and artificial intelligence provides the intelligence necessary to analyze, interpret, and act on that data. The efficiency, responsiveness, and intelligence of connected systems are enhanced by this fusion, which is commonly referred to as the Artificial Intelligence of Things (AIoT). AIoT systems are able to predict, learn from real-time data, and carry out autonomous tasks without the need for human intervention. This combination can be utilized in a variety of fields, including healthcare, agriculture, transportation, smart homes, and industrial automation. Other examples include However, the integration also results in issues with data security, interoperability, and system scalability. The benefits, applications, and repercussions of AI and IoT combined are the focus of this paper's investigation of their synergy.

Index Terms - Real-Time Data, Autonomous Systems, Predictive Analytics, Smart Systems, Healthcare Applications, Industrial Automation, Scalability, Interoperability, Data Security, Future Impact and Artificial Intelligence of Things (AIoT)

Volume 17, Issue 2, April 2026

IJCST Managemet

Current Issue

Recent Issues

Author Guidelines

IJCST RSS Feeds

News & Updates