An Information-Theoretic Approach for Bug Triaging

It is a challenging task to accurately assign bug reports to an expert developer in both Open source project (OSP) as well as closed source projects (CSP). Objective of any triaging techniques is to reduce tossing length thereby minimizing bug fixing time. Developing such an efficient triaging technique is even more challenging for CSP due to limited accessibility of bug reports. As a result, most of the existing triaging techniques have been trained and tested using bugs from OSP leaving CSP untapped to a larger extent. In this paper, we have built a metric based score for identifying and ranking an expert developer independent of the nature of the project. We have proposed two metrics based on information theory i.e. developer specialization and bug importance score. Developer specialization score and Importance score are computed using information gain on bug component and bug importance fields. In order to validate our proposed approach, we choose bug reports from a closed source project developed by company XYZ, pvt. Ltd. India (the real name of the company is not disclosed due to confidentiality reason). Our proposed triaging approach has achieved good prediction results up to 88.9% on the chosen closed source project dataset.