International Research Journal of Education and Technology

Software defects remain one of the most critical challenges in modern software development, with significant implications for reliability, security, and maintenance costs. Traditional fault detection and prediction methods have achieved moderate success but are increasingly limited by the complexity of contemporary software systems. Hybrid machine learning models, which combine multiple learning algorithms and data processing techniques, have emerged as a promising solution to improve both the accuracy and robustness of fault detection and prediction systems. This comprehensive review examines the current landscape of hybrid ML approaches for software fault detection and prediction, analyzing their methodologies, performance metrics, and practical applications. Through systematic examination of 25+ peer-reviewed studies and industry implementations, we identify key trends in ensemble methods, deep learning hybrids, and feature engineering strategies. Our findings indicate that hybrid models consistently outperform single-algorithm approaches, with ensemble methods achieving F1-scores ranging from 0.82 to 0.95 across diverse datasets. However, significant challenges remain in model interpretability, scalability, and real-world deployment. This review concludes by proposing future research directions, including explainable AI integration, transfer learning applications, and lightweight models suitable for edge computing environments. The insights provided aim to guide practitioners and researchers in selecting appropriate hybrid models for specific fault detection and prediction scenarios.

Keywords:fault detection, prediction, hybrid machine