Comprehensive Survey of different Machine Learning Algorithms used for Software Defect Prediction

Abstract

The software development life cycle is a long and complicated process. It consists of analysis, design, development, testing and deployment. Defect prediction is the technique of creating models that detect defective systems such as units or classes in the early stages of the process. The major goal of Software Defect Prediction is to detect defects prone in the program and thereby reduce the effort, time and cost involved to the minimum. This paper gives a comprehensive review of all the techniques to approach defect prediction. The PROMISE repository which is a public software defect prediction dataset which is owned by the National Aeronautics and Space Administration (NASA) is used. More than 30 research papers in the domain of software defect prediction were analysed and reviewed. In each paper surveyed, the processes involved in Software Defect prediction were captured. About 30 papers with different Machine Learning Algorithms were identified and entered into the table. The results were tabulated into columns like Dataset Used, Supervised Learning Algorithm, Un-supervised Learning Algorithm and Computational Intelligence. The bar graphs were generated to determine the most used Supervised Learning Algorithm, Unsupervised Learning Algorithm and Dataset used for Software Defect Prediction among all the 40 papers surveyed thoroughly.