Publication Date: 2024/01/06
Abstract: In the expansive domain of software engineering, the persistent challenge of fault prediction has garnered scholarly interest in machine learning methodologies, aiming to refine decision-making and enhance software quality. This study pioneers advanced fault prediction models, intertwining product and process metrics through machine learning classifiers and ensemble design. The methodological framework involves metric identification, experimentation with machine learning classifiers, and evaluation, considering cost dynamics. Empirically, 42 diverse projects from PROMISE, BUG, and JIRA repositories are examined, revealing advanced models with ensemble methods manifesting an accuracy of (91.7%), showcasing heightened predictive capabilities and nuanced cost sensitivity. Non-parametric tests affirm statistical significance, portraying innovation beyond conventional paradigms. Conclusively, these advanced models navigate inter-project fault prediction with finesse, signifying a convergence of novelty and performance. Simultaneously, anticipating fault proneness in software components is a pivotal focus in software testing. Software coupling and complexity metrics are critical for evaluating software quality. Object-oriented metrics, including inheritance, polymorphism, and encapsulation, influence software quality and offer avenues for estimating fault proneness. This study contributes a comprehensive taxonomy to the discourse, offering a holistic perspective on the multifaceted landscape of object-oriented metrics in fault prediction within the broader context of advancing software quality.
Keywords: Software Fault Prediction; Object-Oriented Testing; Object-Oriented Coupling; Machine Learning, Ensemble Design, Product, and Process Metrics.
DOI: https://doi.org/10.5281/zenodo.10464708
PDF: https://ijirst.demo4.arinfotech.co/assets/upload/files/IJISRT23DEC1931.pdf
REFERENCES