Software Defect Prediction Using Wrapper Feature Selection Based on Dynamic Re-Ranking Strategy

Finding defects early in a software system is crucial task, as it creates adequate time for fixing such using available resources. Strategies symmetric testing have proven useful; however, its inability differentiating incorrect implementations from correct ones drawback. Software defect prediction (SDP) another feasible method that can be used detecting early. Additionally, high dimensionality, data quality problem, has detrimental effect on the predictive capability of SDP models. Feature selection (FS) been solution solving dimensionality issue SDP. According to current literature, two basic forms FS approaches are filter-based feature (FFS) and wrapper-based (WFS). Between two, WFS deemed superior. However, methods computational cost due unknown number executions subset search, evaluation, selection. This characteristic often leads overfitting classifier models easy trapping local maxima. The evaluator maxima overcome by an effective search process. Hence, this study proposes enhanced dynamically iteratively selects features. proposed (EWFS) based incrementally selecting features while considering previously selected space. novelty EWFS enhancement evaluation process deploying dynamic re-ranking strategy germane with low cycle not compromising performance ensuing model. For was deployed Decision Tree (DT) Naïve Bayes classifiers datasets varying granularities. experimental findings revealed outperformed existing metaheuristics sequential search-based established work. fewer less compared methods.