Assessing the Early Bird Heuristic (for Predicting Project Quality)

Before researchers rush to reason across all available data or try complex methods, perhaps it is prudent first check for simpler alternatives. Specifically, if the historical has most information in some small region, a model learned from that region would suffice rest of project. To support this claim, we offer case study with 240 projects, where find those projects “clumps” towards earliest parts A quality prediction just 150 commits works as well, better than state-of-the-art Using ”early bird” data, can build models very quickly and early project life cycle. Moreover, using bird method, have shown simple (with few features) generalizes hundreds projects. Based on experience, doubt prior work generalizing may needlessly complicated an inherently process. Further, focused later-life cycle needs be revisited since their conclusions were drawn relatively uninformative regions. Replication note: our scripts are here: https://github.com/snaraya7/early-bird