Predicting Issues for Resolving in the Next Release

Abstract

Deciding which features or requirements (or commonly referred to as issues) to be implemented for the next release is an important and integral part of any type of incremental development. Existing approaches consider the next release problem as a single or multi-objective optimization problem (on customer values and implementation costs) and thus adopt evolutionary search-based techniques to address it. In this paper, we propose a novel approach to the next release problem by mining historical releases to build a predictive model for recommending if a requirement should be implemented for the next release. Results from our experiments performed on a dataset of 22,400 issues in five large open source projects demonstrate the effectiveness of our approach.