On the classification of bug reports to improve bug localization

Bug localization is the automated process of finding the possible faulty files in a software project. Bug localization allows developers to concentrate on vital files. Information retrieval (IR)-based approaches have been proposed to assist automatically identify software defects by using bug report information. However, some bug reports that are not semantically related to the relevant code are not helpful to IR-based systems. Running an IR-based reporting system can lead to false-positive results. In this paper, we propose a classification model for classifying a bug report as either uninformative or informative. Our approach helps to lower false positives and increase ranking performances by filtering uninformative information before running an IR-based bug location system. The model is based on implicit features learned from bug reports that use neural networks and explicit features defined manually. We test our proposed model on three open-source software projects that contain over 9000 bug reports. The results of the evaluation show that our model enhances the efficiency of a developed IR-based system in the trade-off between precision and recall. For implicit features, our tests with comparisons show that the LSTM network performs better than the CNN and multilayer perceptron with respect to the F-measurements. Combining both implicit and explicit features outperforms using only implicit features. Our classification model helps improve precision in bug localization tasks when precision is considered more important than recall.