Software defect prediction based on nested-stacking and heterogeneous feature selection

Abstract Software testing guarantees the delivery of high-quality software products, and defect prediction (SDP) has become an important part testing. is divided into traditional just-in-time (JIT-SDP). However, most existing frameworks are relatively simplified, which makes it extremely difficult to provide developers with more detailed reference information. To improve effectiveness realize effective resource allocation, this paper proposes a framework based on Nested-Stacking heterogeneous feature selection. The includes three stages: data set preprocessing selection, classifier, model classification performance evaluation. novel selection nested custom classifiers in can effectively accuracy prediction. This conducts experiments two sets (Kamei, PROMISE), demonstrates through comprehensive evaluation indicators, AUC, F1-score. experiment carried out large-scale within-project (WPDP) cross-project (CPDP). results show that proposed excellent types sets, been greatly improved compared baseline models.