Design and Deployment of a Machine Learning Model for Software Defect Detection

This paper explores software defects as an inherent aspect of software products, significantly impacting software quality. Defects, often deviations from specifications, can lead to functionality failures. Ensuring software quality assurance is complex and time-consuming, with many projects lacking sufficient resources to eliminate all defects before release. This can affect product quality and an organization's reputation. To address this challenge, various techniques for software defect prediction are employed. This research utilizes pre-processing, feature extraction, and classification methods. A hybrid approach integrating a genetic algorithm with PSO is used for feature extraction, while bagging classification generates final results. Three ensemble classifiers are implemented, and incorporating PCA for feature reduction further enhances accuracy. Additionally, addressing the class imbalance problem improves prediction performance, achieving 93% accuracy. This study provides an effective approach to software defect detection, contributing to improved software quality and reliability. Key Words: Software Defect, Gaussian Naïve Bayes, Bernoulli Naïve Bayes, Random Forest, PCA, Class Imbalance Handling