A CHANGE GUIDE METHOD BASED ON DEVELOPERS' INTERACTION AND PAST RECOMMENDATION

摘要

In this paper, we propose a change guide method based on the past developers' activity that consists of read and write access records of artifacts. In our proposed method, we calculate candidates of next change recommendation considering the history of its recommendations. We define "cumulative likelihood" to enable the method to recommend the appropriate candidates when a change propagates more than one code elements. A case study using interaction history logs from 15 participants showed the improvement of the accuracy of the method-level change recommendation.