Muhammad Usman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Muhammad Usman has a total of 47 co-authors. Here are the top co-authors by number of shared publications:
- Alvis Cheuk M. Fong
Department of Computer Science, Western Michigan University, Kalamazoo, MI, USA
8 shared publications - Simon Fong
University of Macau, Macao, China
4 shared publications - Muhammad Afzaal
Stockholm University, Stockholm, Sweden
3 shared publications A. C. M. Fong
Western Michigan University
2 shared publications- Azhar Mahmood
Department of Computer Science and Digital Technologies, School of Architecture, Computing, and Engineering, University of East London, London E16 2RD, UK
2 shared publications - Farah Younas
Shaheed Zulfikar Ali Bhutto Inst Sci & Technol, Dept Comp Sci, Islamabad 44000, Pakistan
2 shared publications
Latest Publications
- Open Access
- 12 October 2026
- Published by MDPI AG in Diabetology
- Vol. 7 (10) , 197
A Robust and Explainable Framework for Predicting Hospital Readmission in Patients with Diabetes
Background: Hospital readmission among patients with diabetes remains a healthcare challenge, contributing to adverse clinical outcomes, increased healthcare expenditures, and reduced quality of care. This study presents a leakage-resistant and clinically interpretable machine learning framework for predicting hospital readmission risk using structured clinical and admission-related data. Methods: The framework incorporated preprocessing strategies to minimize information leakage, including duplicate-feature removal, categorical encoding, and feature normalization. Class imbalance was addressed using fold-specific strategies, including class-weighted learning and the Synthetic Minority Over-sampling Technique for Nominal and Continuous features (SMOTENC). Resampling was performed exclusively within each training fold, while validation and test data retained their original class distributions. Feature selection was conducted using correlation analysis and Recursive Feature Elimination (RFE). Principal Component Analysis (PCA) was used for exploratory variance analysis and visualization, preserving the interpretability of the predictive features Five models were evaluated: L2-regularized logistic regression, random forest, ANN, XGBoost, and Hybrid CNN–LSTM. SHAP and LIME provided global and local explanations. Results: The hybrid CNN-LSTM achieved the strongest performance, attaining an accuracy of 0.80 and an area under the receiver operating characteristic curve (ROC-AUC) of 0.86 on a hold-out dataset. Explainability analyses enhanced model transparency and supported clinical interpretation of predictions. Conclusions: Combining leakage-resistant preprocessing, balanced training, interpretable feature selection, and explainable deep learning provides a robust and clinically meaningful framework for diabetic readmission prediction.
- 12 October 2026
- Published by MDPI AG in Diabetology
- Vol. 7 (10) , 197
Open AccessA Robust and Explainable Framework for Predicting Hospital Readmission in Patients with Diabetes
Background: Hospital readmission among patients with diabetes remains a healthcare challenge, contributing to adverse clinical outcomes, increased healthcare expenditures, and reduced quality of care. This study presents a leakage-resistant and clinically interpretable machine learning framework for predicting hospital readmission risk using structured clinical and admission-related data. Methods: The framework incorporated preprocessing strategies to minimize information leakage, including duplicate-feature removal, categorical encoding, and feature normalization. Class imbalance was addressed using fold-specific strategies, including class-weighted learning and the Synthetic Minority Over-sampling Technique for Nominal and Continuous features (SMOTENC). Resampling was performed exclusively within each training fold, while validation and test data retained their original class distributions. Feature selection was conducted using correlation analysis and Recursive Feature Elimination (RFE). Principal Component Analysis (PCA) was used for exploratory variance analysis and visualization, preserving the interpretability of the predictive features Five models were evaluated: L2-regularized logistic regression, random forest, ANN, XGBoost, and Hybrid CNN–LSTM. SHAP and LIME provided global and local explanations. Results: The hybrid CNN-LSTM achieved the strongest performance, attaining an accuracy of 0.80 and an area under the receiver operating characteristic curve (ROC-AUC) of 0.86 on a hold-out dataset. Explainability analyses enhanced model transparency and supported clinical interpretation of predictions. Conclusions: Combining leakage-resistant preprocessing, balanced training, interpretable feature selection, and explainable deep learning provides a robust and clinically meaningful framework for diabetic readmission prediction.
- 8 August 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 351
SWARM: A novel Sample Weighting Approach for Rare Minority data classification in imbalanced and evolving data streams
- 8 August 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 351
SWARM: A novel Sample Weighting Approach for Rare Minority data classification in imbalanced and evolving data streams
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 126182-126198
An Asynchronous Gradient Fusion Framework for Distributed Deep Learning in Tuberculosis Chest X-Ray Diagnosis
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 126182-126198
Open AccessAn Asynchronous Gradient Fusion Framework for Distributed Deep Learning in Tuberculosis Chest X-Ray Diagnosis
- 26 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Consumer Electronics Magazine
- Vol. 15 (2) , 47-57
Multilabel Learning for Massive Categorization of Resources in Metaverse across SmartX Environments
The much-anticipated metaverse will fundamentally transform how consumers interact with machines across SmartX environments. Information overload is a problem that metaverse users can encounter when they navigate a myriad of multimedia resources. Providing effective and timely categorization of resources can facilitate effective human–human communications in metaverse. Existing categorization techniques for the current web have four key limitations: segregation of feature selection and feature correlation, validation using small datasets, lack of comparative analysis of multilabel classifiers, and lack of a framework for selecting optimal classifiers. These limitations hinder existing methods’ applicability to metaverse. In response, this article presents an effective framework for multilabel categorization of metaverse resources using natural language processing and machine learning. The proposed framework comprises several major steps, including extraction of contents, preprocessing, removing redundant and ambiguous data, converting text into features, and then feature selection and classification. The proposed framework has been implemented as a predictive model for metaverse users and tested on benchmark and real-world datasets. Experimental evaluation is conducted based on multiple parameters: accuracy, average precision, Hamming loss, one error, and ranking loss. Experimental results show a maximum increase in accuracy of 10% for both datasets. Comparative analysis indicates the proposed framework outperforms existing models not only by accuracy but also with advanced evaluation measures, such as Hamming loss, ranking loss, and one error.
- 26 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Consumer Electronics Magazine
- Vol. 15 (2) , 47-57
Multilabel Learning for Massive Categorization of Resources in Metaverse across SmartX Environments
The much-anticipated metaverse will fundamentally transform how consumers interact with machines across SmartX environments. Information overload is a problem that metaverse users can encounter when they navigate a myriad of multimedia resources. Providing effective and timely categorization of resources can facilitate effective human–human communications in metaverse. Existing categorization techniques for the current web have four key limitations: segregation of feature selection and feature correlation, validation using small datasets, lack of comparative analysis of multilabel classifiers, and lack of a framework for selecting optimal classifiers. These limitations hinder existing methods’ applicability to metaverse. In response, this article presents an effective framework for multilabel categorization of metaverse resources using natural language processing and machine learning. The proposed framework comprises several major steps, including extraction of contents, preprocessing, removing redundant and ambiguous data, converting text into features, and then feature selection and classification. The proposed framework has been implemented as a predictive model for metaverse users and tested on benchmark and real-world datasets. Experimental evaluation is conducted based on multiple parameters: accuracy, average precision, Hamming loss, one error, and ranking loss. Experimental results show a maximum increase in accuracy of 10% for both datasets. Comparative analysis indicates the proposed framework outperforms existing models not only by accuracy but also with advanced evaluation measures, such as Hamming loss, ranking loss, and one error.
- 30 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Engineering Management Review
- Vol. 53 (1) , 63-72
Public Health Management for Maternal Mortality Reduction in Smart Cities
More than half a million women die each year from pregnancy-related causes, with 99% of these deaths occurring in developing countries. Rapid development of smart cities in some of these developing countries, coupled with technological advances, offers an opportunity for better management of maternal healthcare. Using Sindh, Pakistan as a case study, we present an automated system for early prediction of maternal mortality rate (MMR). Previous prediction models were not effective due to a lack of integration of healthcare program datasets to identify the variables influencing MMR across healthcare programs. The novel system was validated through an integrated approach, applying different machine learning algorithms to MMR datasets taken from Sindh healthcare programs. With prediction accuracies >90%, Public health officials now havse an easy-to-use tool to gain insight into factors that influence maternal maternity, which can lead to MMR reduction. Throughout the study, we have ensured that all research steps and findings can be broadly applicable to other smart city environments.
- 30 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Engineering Management Review
- Vol. 53 (1) , 63-72
Public Health Management for Maternal Mortality Reduction in Smart Cities
More than half a million women die each year from pregnancy-related causes, with 99% of these deaths occurring in developing countries. Rapid development of smart cities in some of these developing countries, coupled with technological advances, offers an opportunity for better management of maternal healthcare. Using Sindh, Pakistan as a case study, we present an automated system for early prediction of maternal mortality rate (MMR). Previous prediction models were not effective due to a lack of integration of healthcare program datasets to identify the variables influencing MMR across healthcare programs. The novel system was validated through an integrated approach, applying different machine learning algorithms to MMR datasets taken from Sindh healthcare programs. With prediction accuracies >90%, Public health officials now havse an easy-to-use tool to gain insight into factors that influence maternal maternity, which can lead to MMR reduction. Throughout the study, we have ensured that all research steps and findings can be broadly applicable to other smart city environments.
- 19 June 2023
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 83 (3) , 6883-6908
Deep learning framework for early detection of COVID-19 using X-ray images
- 19 June 2023
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 83 (3) , 6883-6908
Deep learning framework for early detection of COVID-19 using X-ray images
- 8 October 2022
- Published by Springer Nature in Journal of Ambient Intelligence and Humanized Computing
- Vol. 14 (7) , 9233-9251
Highly accurate phishing URL detection based on machine learning
- 8 October 2022
- Published by Springer Nature in Journal of Ambient Intelligence and Humanized Computing
- Vol. 14 (7) , 9233-9251
Highly accurate phishing URL detection based on machine learning
- Open Access
- 1 March 2022
- Published by Elsevier in Egyptian Informatics Journal
- Vol. 23 (1) , 137-147
A Hierarchical Energy Conservation Framework (HECF) of Wireless Sensor Networks by Temporal Association Rule Mining for Smart Buildings
- 1 March 2022
- Published by Elsevier in Egyptian Informatics Journal
- Vol. 23 (1) , 137-147
Open AccessA Hierarchical Energy Conservation Framework (HECF) of Wireless Sensor Networks by Temporal Association Rule Mining for Smart Buildings
- 7 August 2021
- Published by Springer Nature
- p. 517-535
A Critical Evaluation of Machine Learning and Deep Learning Techniques for COVID-19 Prediction
- 7 August 2021
- Published by Springer Nature
- p. 517-535
A Critical Evaluation of Machine Learning and Deep Learning Techniques for COVID-19 Prediction