Improving software quality based on relationship among the change proneness and object oriented metrics

With the demand of increasing functionality and arrival of defects, software goes through a lot of changes therefore its quite challenging task to maintain the quality of the software. In this paper we developed models to predict the change proneness of the classes in the object oriented system by analyzing the relationship between the object oriented metrics and change proneness. The model proposed is also validated by object oriented open source software. We have analyzed our results by the Receiver Operator Characteristics Curve. The results thus obtained shows that there is a significance relationship between the object oriented metrics and change proneness. We have analyzed statistical as well as machine learning techniques and the results shows that machine learning techniques are the good predictors of the change proneness. Rigorous testing of these change prone classes may improve the quality of the software and it may also reduce our work at the maintenance phase.