GAP: Forecasting commit activity in git projects

•Assess the risk of contributor abandonment by identifying declining activity early.•Predict the activity of individual contributors based on recent commit activity.•The model accuracy is validated on 17,340 pairs of contributors and projects.•An open source command-line tool GAP implements the prob...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:The Journal of systems and software Ročník 165; s. 110573
Hlavní autori: Decan, Alexandre, Constantinou, Eleni, Mens, Tom, Rocha, Henrique
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Elsevier Inc 01.07.2020
Predmet:
Commit activity
Developer abandonment
Distributed software development
Git
Prediction model
Distributed software development
Commit activity
Git
Developer abandonment
Prediction model
ISSN:0164-1212, 1873-1228
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:•Assess the risk of contributor abandonment by identifying declining activity early.•Predict the activity of individual contributors based on recent commit activity.•The model accuracy is validated on 17,340 pairs of contributors and projects.•An open source command-line tool GAP implements the probabilistic forecasting model. Abandonment of active developers poses a significant risk for many open source software projects. This risk can be reduced by forecasting the future activity of contributors involved in such projects. Focusing on the commit activity of individuals involved in git repositories, this paper proposes a practicable probabilistic forecasting model based on the statistical technique of survival analysis. The model is empirically validated on a wide variety of projects accounting for 7528 git repositories and 5947 active contributors. We found that a model based on the last 20 observed days of commit activity per contributor provides the best concordance. We also found that the predictions provided by the model are generally close to actual observations, with slight underestimations for low probability predictions and slight overestimations for higher probability predictions. This model is implemented as part of an open source tool, called GAP, that predicts future commit activity.
ISSN:0164-1212
1873-1228
DOI:10.1016/j.jss.2020.110573