Developer Behavior and Sentiment from Data Mining Open Source Repositories
William N. Robinson, Tianjie Deng, Zirun Qi · 2016
Developer sentiment may wax and wane as a project progresses. Open-source projects that attract and retain developers tend to be successful. It may be possible to predict project success, in part, if one can measure developer behavior and sentiment -- projects with active, happy developers are more likely to succeed. We have analyzed GitHub.com projects in an attempt to model these concepts. We have data mined 124 projects from GitHub.com. The projects were automatically mined using sequence mining methods to derive a behavioral model of developer activities. The projects were also mined for developer sentiment. Finally, a regression model shows how sentiment varies with behavioral differences -- a change in behavior is correlated with a change in sentiment. The relationship between sentiment and success is not directly explored, herein. This research project is a preliminary step in a larger research project aimed at understanding and monitoring FLOSS projects using a process modeling approach.