A Hybrid Fault-Proneness Detection Approach Using Text Filtering and Static Code Analysis

We have proposed a fault-prone software module detection method using text-filtering approach, called Fault-proneness filtering. Even though the fault-proneness filtering achieved high accuracy in detecting fault-prone modules, the detail of each fault cannot be specified enough. We thus try to complete such weakness of the fault-proneness filtering by using static code analysis. To do so, we analyze characteristics of fault-proneness filtering and a static code analyzer, PMD, by applying both methods to open source software projects. The result of comparison tells us that faultproneness filtering can capture similar faults related to “braces” and “code size” rules of PMD. Furthermore, fault-proneness filtering can reduce false positives of rules with high false positive rate such as “design”, “naming”, and “optimization”. According to the results of analysis, we can thus construct a hybrid fault-proneness detection method using fault-proneness filtering and PMD.