[研究発表] Can Developers’ Interaction Data Improve Change Recommendation?
- 07 /
小林研OBの山森 章弘さんが、イタリアのトリノで開催された IEEE 41st Annual Computer Software and Applications Conference (COMPSAC 2017) にて研究成果を発表しました。
著者： Akihiro Yamamori, Anders Mikael Hagward, and Takashi Kobayashi (Tokyo Tech)
題目： Can Developers’ Interaction Data Improve Change Recommendation?
掲載誌： in Proc. COMPSAC’17
One of the most common causes of bugs is overlooking changes. To prevent bugs and improve the quality of the products, numerous studies have been undertaken on change guides based on logical couplings extracted from developers’ past process histories, such as change history. While valuable change rules based on logical couplings can be gleaned found from the change history, these rules often fail to find appropriate candidates because the change histories in repositories only preserve a summary of changes between commits. We recently analyzed the interaction data produced by a developer in an integrated development environment. Such interaction data contains not only a detailed change history but also reference activities between commits. In this paper, we investigate whether logical couplings extracted from interaction data could improve change recommendation performance. We used the interaction data from actual open source development, not from the project only for this study. Experimental results obtained using the interaction data from actual open source development showed a significant improvement in the efficiency of the change recommendation process. The results also indicated improvement in the number of detected artifacts that the developer had forgot to change.