Authors:
Suyash Shukla
and
Sandeep Kumar
Affiliation:
Department of Computer Science and Engineering, Indian Institute of Technology Roorkee, Roorkee, India
Keyword(s):
Issue Tracking System, Machine Learning, Term Frequency Inverse Document Frequency, Smartshark.
Abstract:
Every software deals with issues such as bugs, defect tracking, task management, development issue to a customer query, etc., in its entire lifecycle. An issue-tracking system (ITS) tracks issues and manages software development tasks. However, it has been noted that the inferred issue types often mismatch with the issue title and description. Recent studies showed machine learning (ML) based issue type prediction as a promising direction, mitigating manual issue type assignment problems. This work proposes an ensemble method for issue-type prediction using different ML classifiers. The effectiveness of the proposed model is evaluated over the 40302 manually validated issues of thirty-eight java projects from the SmartSHARK data repository, which has not been done earlier. The textual description of an issue is used as input to the classification model for predicting the type of issue. We employed the term frequency-inverse document frequency (TF-IDF) method to convert textual descri
ptions of issues into numerical features. We have compared the proposed approach with other widely used ensemble approaches and found that the proposed approach outperforms the other ensemble approaches with an accuracy of 81.41%. Further, we have compared the proposed approach with existing issue-type prediction models in the literature. The results show that the proposed approach performed better than existing models in the literature.
(More)