MACHINE LEARNING TECHNIQUES FOR PREDICTING SOFTWARE DEFECTS IN LARGE CODEBASES

Paper ID: EIJTEM_2025_12_2_27-33

Author's Name: Aniket Kulkarni

Volume: 12

Issue: 2

Year: 2025

Page No: 27-33

Abstract:

Software defects can significantly impact the quality, reliability, and security of software systems. As codebases grow in size and complexity, traditional manual testing and static analysis techniques become less effective. Machine learning (ML) techniques provide a powerful approach to predicting software defects by analyzing historical code changes, bug reports, and software metrics. This paper explores various ML techniques, including decision trees, support vector machines, neural networks, and ensemble methods, for defect prediction in large-scale software projects. We discuss the effectiveness of feature selection, data preprocessing, and model evaluation strategies to enhance prediction accuracy. Furthermore, we highlight challenges such as class imbalance, explainability, and integration into software development workflows. Experimental results demonstrate that ML-based defect prediction can improve defect detection rates, optimize testing efforts, and enhance software quality assurance.

Keywords: Machine learning, software defect prediction, large codebases, software quality, defect detection, predictive analytics, neural networks, decision trees, software metrics, automation.

View PDF