Mining packages change history to predict their future change-coupling behavior

In development and maintenance phases, packages of object oriented (OO) software applications are changed. During change process, change in one package in most of the cases will result in changes in other packages. The accurate prediction of future change-coupling behavior provides a significant challenge for development or maintenance team. In this paper, we have used association mining for analyzing the change history of the packages of an OO application and proposed an approach based on mining results and graph clustering for predicting change coupling. Firstly, a framework has been proposed to measure changeability index and change readiness of packages as well as frequently changed packages of an application. Secondly, package clustering has been done to generate clusters of packages that are highly change-coupled. A case study on a Java application has been done to demonstrate our work as well as to analyze the results. The results obtained from this study show that the proposed approach can be an essential aid for development team to predict future change-coupling behavior of packages.