Cross Project Software Defect Prediction Using Machine Learning: A Review

Abstract

Software defect prediction is a crucial area of study focused on enhancing software quality and cutting down on software upkeep expenses. Cross Project Defect Prediction (CPDP) is a method meant to use information from different source projects to spot software issues in a specific project. CPDP comes in handy when the project being analyzed lacks enough or any data about defects for creating a dependable defect prediction model. Machine learning that is a part of artificial intelligence learns from data and then makes forecasts or choices. Machine learning (ML) is a key component of CPDP because it can learn from heterogeneous and imbalanced data sources. However, there are many challenges and open issues in applying machine learning to CPDP, such as data selection, feature extraction, model selection, evaluation metrics, and transfer learning. In this study, we provide a complete review of existing literature from 2018 to 2023 on Defect Prediction using Machine Learning, covering the main methods, applications, and limitations. We also use ML to identify current research gaps and future directions for CPDP. This paper will serve as a useful reference for researchers interested in using ML for CPDP.