Digging for diamonds: Identifying valuable end-user code in repositories

To a large extent, repositories of end-user code are “write-only”: much of the code that people publish never sees substantial reuse. Yet buried within these repositories are valuable pieces of code, though finding them is not always easy. In prior work, we developed a model that can predict, when a web macro is created, whether that script will be reused by anybody. In the current paper, we analyze data from two other end-user repositories to investigate the model’s generalizability to other kinds of code. We find that the model performs well for a wide range of different purposes and configurations, including predicting future reuse events based on data about past events, indicating that the model could serve as an effective basis for future repository enhancements.