Ensembled Fault-Prone Software Module Prediction Model for Imbalanced Dataset

Software defect prediction (SDP) is an assessment that enables stakeholders to estimate its quality, reliability, functionality, scalability, information security, and availability during SDLC of software. With the digitalization of business and processes, its scope has multiplied since business and process owners look for reliable solutions. It is an area of research in software engineering. It can be done manually or automatically. Researchers have been trying to predict defects in a software module by applying various techniques to automatically detect defects in software modules. AI techniques like machine learning (ML), deep learning (DL), etc. have been used extensively by researchers to find defects in software automatically. A software module can be defective/non-defective, bug/no bug, hence most existing methods regard SDP problems under classification-based problems. In this study, we have developed a classification-based SDP model by applying well-established machine learning classifiers like randon forest (RF), XGBoost, AdaBoost, SVM, ANN, etc. on open-source projects to predict module-level defects in software. The results suggest that ANN, RF, XGBoost, AdaBoost and SVM-based classifiers significantly improve the prediction accuracy of software defects compared to previous models (Cetiner and Sahingoz, A comparative analysis for machine learning based software defect prediction systems, pp. 1–7. IEEE, 2020).