Predicting components for issue reports using deep learning with information retrieval

Abstract

Assigning an issue to the correct component(s) is challenging, especially for large-scale projects which have are up to hundreds of components. We propose a prediction model which learns from historical issues reports and recommends the most relevant components for new issues. Our model uses the deep learning Long Short-Term Memory to automatically learns semantic features representing an issue report, and combines them with the traditional textual similarity features. An extensive evaluation on 142,025 issues from 11 large projects shows our approach outperforms alternative techniques with an average 60% improvement in predictive performance.