Automation of Bug-Report Allocation to Developer using a Deep Learning Algorithm

Software bug maintenance is an important aspect of all software projects. The assignment of bug reports is essential in order to resolve bugs efficiently. In the case of open-source software developments and large projects, where many developers are engaged on different aspects of software development, it can be difficult to assign bug removal tasks to an appropriate developer. An increase in reported bugs, coupled with an increase in the number of software developers, will complicate the bug triage process. In these situations, bug triaging might be slow and may increase the Bug Tossing Length (BTL). An automated system to triage bug reports could potentially reduce BTL, as manual assignment of bug reports is tiresome, costly, and very time-consuming. The assignment of bug reporting to an irrelevant developer who does not possess sufficient experience to deal with the bug will adversely impact BTL and customer satisfaction. Text-based classification methods have the potential to make a strong contribution to automating the bug triaging process. In this research, different types of Information Retrieval and Machine Learning algorithms are used to determine the appropriate developer/s to rectify the reported bugs. This study used deep learning algorithms, such as the Bidirectional Long Short-Term Memory Network, to automate the bug triaging process. Bug reports contain textual data related to the bug information. In this research, the pretrained GloVe model is employed for word-to-vector representation of bug reports’ textual information. In this framework, developers’ activities are monitored based on their working history. To test the proposed approach, three large datasets, Net-Beans, Eclipse, and Mozilla, are used. It was observed that the proposed technique produced better results in terms of accuracy, recall, precision and f-measure compared to traditional Machine Learning algorithms for bug report recommendation.