Object Oriented Fault Prediction Analysis Using Machine Learning Algorithms

摘要

One of the important key element in the development and maintenance process of software is fault prediction, it concerns the overall success of the system. Predicting software faults at the initial phase helps the developer to build reliable software and also minimise the cost. Fault prediction model provide insights to development team about faulty behaviour and thus act accordingly. The study presented in this paper discusses the performances of various machine learning algorithms in predicting fault prone classes and also investigates the role played by different software metrics of the datasets. First we apply the correlation based feature selection technique to get set of uncorrelated metrics that are highly desirable and informative for prediction. Then we develop model for prediction with the help of some supervised machine learning techniques. These models are validated on six different versions of object oriented java project obtained from GitHub.