Prediction and recommendation models of software ecosystem health

Description

Highly collaborative software ecosystems are omnipresent. They include, for example, open source software repositories, software package distributions, software libraries, software product lines, social networks and collaborating software communities (e.g., StackExchange). The sustainability and survival of such collaborative software ecosystems can depend on a large number of factors that determine the ecosystem’s health. Some examples are its resilience, i.e., the ability to withstand internal changes or external perturbations; its social, technical or organizational diversity; the presence of security vulnerabilities and bugs; the risk of key developers abandoning the ecosystem; the presence of technical debt and community smells, and many more. This research aims to take a macroscopic socio-technical view on the study of such health issues in large interdependent collections of software components maintained by interacting communities of contributors. The goal is to define a suite of metrics and corresponding prediction models of decreasing health, and propose automated recommendations of ways to improve the health. predict the quality and survival of projects or contributors in their ecosystems, to propose and support concrete guidelines and actions to improve this survival, and to validate these guidelines in practice. To do so, we will rely on techniques and ideas borrowed from the domains of software repository mining, empirical software engineering and data analytics, but also borrow ides from ecology, statistics and economical models.The proposed topic is related to the ongoing interdisciplinary and interuniversity research project SECOHealth www.secohealth.org