Towards Issue Recommendation for Open Source Communities

Abstract

In open source software development, a major challenge is the prioritization of new requirements as well as the identification of responsible developers for their implementation. Unlike conventional industrial software development, where requirements engineers have to explicitly define who implements what, in the context of open source development, developers (contributors) usually decide on their own which requirements to implement next. Contributors have to deal with a huge number of requirements where the recognition of the most relevant ones often becomes a crucial task with a high impact on the success of a software project. This fact defines our major motivation for the development of a prioritization tool for the ECLIPSE community which recommends relevant requirements (issues/bugs) to open source developers. Our tool uses real-world data from ECLIPSE in order to build a prediction model. We trained and tested our tool with different classifiers such as Naive Bayes (representing our baseline), Decision Tree, and Random Forest. The evaluation results indicate that the Random Forest classifier correctly predicts issues with a precision of 0.88 (F1-score 0.68).CCS CONCEPTS• Information systems → Recommender systems; • Humancentered computing → Open source software; • Computing methodologies → Machine learning approaches; Natural language processing; • Software and its engineering → Requirements analysis.