期刊名称：Software Engineering : an International Journal
电子版ISSN：2249-9342
出版年度：2011
卷号：1
期号：1
页码：24-37
出版社：Delhi Technological Universiity
摘要：Software testing is a critical and essential part of software development that consumes maximum resources and effort. The construction of models to predict faulty classes can help and guide the testing community in predicting faulty classes in early phases of software development. It is important to analyze and compare the predictive accuracy of machine learning classifiers. The aim of this paper is to find the relation of object oriented metrics and fault proneness of a class. We have used seven machine learning and one logistic regression method in order to predict faulty classes. The results of our work are based on data set obtained from open source software. The results show that the predictive accuracy of machine learning technique LogitBoost is highest with AUC of 0.806.
