A Detailed Literature Survey and Analysis of Heterogeneous Cross-Project Defect Prediction

Abstract

To improve the quality of software and efficiency of the testing process, identification of the defect prone modules at right time is known as software defect prediction. In recent years, the researchers started focusing on automated software defect prediction using cross project previous data instead of same project or with identical measures. This increases the capability of software defect prediction in an optimized manner. But handling the heterogenous cross project defect prediction (HCPDC) is considered as toughest challenge because of the gap between source and target projects. This paper performs a depth analysis of the various existing literatures related to the prediction models of HCPDC and CPDC using machine learning and statistical methods. Finally, the primary challenges in selection of metrics, class imbalance handling and distribution of the datasets are discussed.