Automated Identification of Change-Prone Classes in Open Source Software Projects

Abstract

Identifying change-prone classes can enable developers to pay more attention to classes with similar characteristics in the future and thus test resources and time can be used more effectively. In this paper, we collect a set of static metrics and change data at class level from an open-source software product, Datacrow. With this data, we first validate Pareto's Law and find that about 80% of the lines changed are located in only 20% of the classes. We then use classification methods to identify these change-prone classes. Our experimental results show that our classification results are useful for identifying change-prone classes and thus can help to improve the efficiency of developers.