Software Defect Prediction Using Weighted Feature Selection

Abstract

Software defect prediction (SDP) is essential for maintaining software quality and minimizing maintenance costs by identifying defect-prone modules early in the development lifecycle. The success of SDP models largely depends on the selection of relevant features that contribute significantly to prediction accuracy. This research introduces a novel approach to SDP using weighted feature selection, where features are assigned weights based on their relevance to the defect prediction task. The proposed method combines statistical and model-based techniques to prioritize impactful features, leading to enhanced model performance. Extensive experiments are conducted on datasets from the PROMISE repository, demonstrating significant improvements in accuracy, precision, recall, F1-score, and area under the ROC curve (AUC) compared to traditional feature selection methods. The findings suggest that weighted feature selection not only improves defect prediction accuracy but also enhances model interpretability. The study's implications extend beyond defect prediction, offering potential applications in various machine learning tasks.