Forecasting OSS Sustainability

# Description

This work discusses how more than 80% of OSS projects fail and the importance of identifying factors associated with success to devise interventions when a project is failing. The success of OSS has been studied through various angles, including empirical studies of diverse projects using productivity and community popularity as proxies for sustainability. The paper argues that the internal socio-technical structure of projects is also important for sustainability, which highlights the importance of self-sustaining and self-governing communities. To understand how nascent projects evolve into sustainable ones, we applied socio-technical network modeling to a dataset of Apache Software Foundation Incubator projects and developed interpretable models that can forecast a project's sustainability with 93+% accuracy. We propose a strategy for real-time monitoring and suggesting actions that can be used by projects to correct their sustainability trajectories.

# Findings

• We propose the first OSS project sustainability forecast measure modeled from tens of socio-technical network and project features. Our model shows excellent predictive performance (> 93% accuracy as early as 8 months into incubation).
• We find that ASF incubator projects with fewer but more focused committers and more but distributed (participating in asynchronous discussions) communicators are more likely to gain momentum to self-sustainability.
• We describe a strategy for real-time monitoring of the sustainability forecast for any project, derived from an interpretable version of our DNN model.

# Paper

The paper can be found here