An AST-Based Approach to Classifying Defects

During software development life cycle, a large number of defects could be discovered and reported. Defect classification is essential, as it can help developers better understand the distribution of the defect types and hence help reduce the effort of root cause analysis. In this paper, we proposed an approach to automatically classifying software defects using various features extracted from the Abstract Syntax Tree (AST) of the source code. We evaluated our approach by classifying 1174 defects from MapReduce, Tomcat, and Solr. The experimental results show that the proposed approach can predict interface defects and control/logic defects well.