ASE 2017

2017 32nd IEEE/ACM International Conference on Automated Software Engineering (ASE 2017), October 30 – November 3, 2017, Urbana-Champaign, IL, USA

Desktop Layout

Recommender Systems
Technical Research
Predicting Relevance of Change Recommendations
Thomas Rolfsnes, Leon Moonen, and David Binkley
(Simula Research Laboratory, Norway; Loyola University Maryland, USA)
Supplementary Material
Abstract: Software change recommendation seeks to suggest artifacts (e.g., files or methods) that are related to changes made by a developer, and thus identifies possible omissions or next steps. While one obvious challenge for recommender systems is to produce emph{accurate} recommendations, a complimentary challenge is to emph{rank} recommendations based on their emph{relevance}. In this paper, we address this challenge for recommendation systems that are based on emph{evolutionary coupling}. Such systems use emph{targeted association-rule mining} to identify emph{relevant patterns} in a software system's change history. Traditionally, this process involves ranking artifacts using emph{interestingness measures} such as emph{confidence} and emph{support}. However, these measures often fall short when used to assess recommendation relevance. We propose the use of random forest classification models to assess recommendation relevance. This approach improves on past use of various interestingness measures by learning from previous change recommendations. We empirically evaluate our approach on fourteen open source systems and two systems from our industry partners. Furthermore, we consider complimenting two mining algorithms: textsc{Co-Change} and textsc{Tarmaq}. The results find that random forest classification significantly outperforms previous approaches, receives lower Brier scores, and has superior trade-off between precision and recall. The results are consistent across software system and mining algorithm.


Time stamp: 2020-04-05T05:19:10+02:00