[研究発表] Effects of Recency and Commits Aggregation on Change Guide Method Based on Change History Analysis

小林研M1の森さんが，11/15-20にスペイン・バルセロナで開催されたThe Tenth International Conference on Software Engineering Advances (ICSEA2015)にて，自身の卒業論文を発展させた研究成果を発表いたしました．

著者： Tatsuya Mori, Anders Mikael Hagward, Takashi Kobayashi
題目：Effects of Recency and Commits Aggregation on Change Guide Method Based on Change History Analysis
掲載誌：Proc. the Tenth International Conference on Software Engineering Advances (ICSEA2015) pp.96-101
概要：
Many studies on change guide, which support to suggest necessary code changes with using co-change rules extracted from a change history, have been performed. The most important aim of our study is prevention of overlooked changes, i.e., we aim to recommend developers to files which they should change but have not been done yet at the point when they decide to commit. The recommendations by existing tools are adequately accurate. However, the scope of the overlooked changes that those tools can recommend is small. In this study, we focus on two methods to recommend more overlooked changes. One is the consideration of recency, i.e., we use only recent commits for extracting co-change rules. The other is the aggregation of commits for the same task, i.e., we aggregate consecutive commits assigned a same bug. We investigated how the consideration of recency and the aggregation of consecutive commits assigned a same bug affect the quality of co-change rules. Experimental results using typical OSS showed that the consideration of recency can improve the recommendation performance. We could extract more useful co-change rules with the consideration of recency. We could also recommend overlooked changes in a higher rank than without the consideration of recency. Regarding the aggregation of consecutive commits assigned a same bug, we found that the performance of recommendations can be improved depending on projects