Software Bug Prediction Employing Feature Selection and Deep Learning

It was proven that the cost of fixing errors escalates as a project moves through its life cycle in an exponential fashion. Identifying buggy classes, as soon as they are committed to the Version Control System, would have a significant impact on reducing such cost. Mining in software repositories is a growing research area, where innovative techniques and models are designed to analyze software repositories data and uncover useful information that can help in software bug prediction. Previous studies showed that Deep Learning has achieved remarkable results in many fields and it keeps evolving. In this paper, experiments are carried out to study the effect of feature selection on the performance of bug prediction models and to check if better results can be obtained by using the promising Deep Learning techniques. Results show that applying feature selection, using a simple filter approach, such as selecting the highly ranked 9 and 5 features out of the 17 features, did not enhance the performance measures in most cases. On the other hand, results show that Deep Learning model (DL) achieves higher performance measures than the selected set of base classifiers for small and balanced datasets.