Interpretable Test Case Recommendation based on Knowledge Graph

Abstract

Reproducing bugs and identifying causes is essential for the debugging of complex software systems. However, existing test case selection and recommendation technique diagnose bugs but failed to provide information to understand the cause. In this paper, we present an interpretable test case recommendation technique by building up knowledge graphs based on massive test cases, bug reports, code changes, and documents stored in software repositories. Specifically, it identifies correlations between new issue reports and historical information based on the knowledge graph and thus present test cases and corresponding documents to support the bug diagnosis. We conduct an empirical study on autonomous driving systems to show our technique is capable of identifying the proper test case. Further, we validate the effectiveness of recommended interpretation. The study shows that the recommended interpretation can help testers to comprehend bug reports and diagnose bugs efficiently.