Improving Software Defect Prediction Performance through Effective Forward Selection Method with Linear Regression for Handling Class Imbalance

In software engineering, accurately predicting software defects is critical for maintaining software quality and reducing development costs. This research aims to improve the performance of software defect prediction models by employing forward selection with linear regression on imbalanced datasets. We utilized 12 datasets from the NASA Metrics Data Program (MDP) repository to comprehensively evaluate the impact of feature selection on prediction accuracy and the Area Under the Curve (AUC). The results showed that the model with feature selection has a higher average accuracy (91.11%) than without (78.67%). Additionally, the AUC increased from an average of 75.33% to 81.12% following feature selection. The paired t-test indicates that this difference is statistically significant $(t(11)=-3.2017, p\lt0.05)$. These findings emphasize the importance of feature selection in improving model performance. The improvement in accuracy and AUC highlights the effectiveness of forward selection in enhancing the predictive performance of linear regression models, especially when dealing with imbalanced data. These insights provide valuable guidance for software quality assurance practices, suggesting that integrating forward selection with linear regression can lead to more reliable defect prediction outcomes.