Development of Models for Improving Software Maintenance

Abstract

With the emerging technology change now a days softwares are going through a lot of change resulting in development of so many versions and hence it is quite difficult to maintain the quality of the software. In this project we will develop models by applying machine learning and statistical techniques on the object oriented metrics to find out the change prone classes .We find relationship between object-oriented metrics given by Chidamber and Kemerers and change proneness. The results are validated using open source softwares. We will also compare and assess various machine learning techniques for predicting change in a class.The performance of the predicted models was evaluated using receiver operating characteristic analysis. By assessing the results we found out that machine learning techniques are better in comparison to statistical techniques .Another analysis showed that models constructed for a software can also be used to predict change proneness of classes of other software in same domain.