An Approach for Predicting Bug Report Fields Using a Neural Network Learning Model
Korosh Koochekian Sabor, Mathieu Nayrolles, Abdelaziz Trabelsi, Abdelwahab Hamou‐Lhadj · 2018
Bug fixing is a major activity in software development and maintenance. Developers use information reported in bug reports to identify bug causes and to provide a fix. Previous studies have shown that bug report fields are often reassigned to different development teams after the report is sent to developers. This can introduce considerable time delays in the bug handling process. To overcome this issue, there exist numerous methods to automatically predict bug report fields by examining historical data. In this paper, we combine a neural network based model with stack traces to predict bug report fields. When applied to bug reports of the Eclipse repository, we found that our approach achieves improvement of about 73% in product prediction accuracy, and 85% in component prediction accuracy over the well-known K-nearest neighbors (KNN) algorithm. The results of this preliminary study suggest that neural networks provide a robust alternative to traditional classification techniques.