Feature-Based Fault Localization in Evolving Software: Leveraging Regression Testing Insights

Fault localization remains a vital yet resource-intensive task, particularly within software evolution, where swift and accurate fault localization is crucial. Whereas substantial research has improved fault localization techniques, challenges persist specifically in evolving software systems. Simultaneously, regression testing has been investigated to enhance testing efficiency while maintaining fault localization accuracy. Despite these advancements, further methodological improvements are necessary to refine fault localization accuracy in evolving software environments. This study proposes a new combination of features extracted from the source code and the associated test suite. The features include coverage, data and control dependencies, and code changes, with a new feature, “Test Case Weight,” derived from the test suite. “Test Case Weight” incorporates the effects of regression testing, which is influenced by two regression testing techniques: test case reduction and selection. Consequently, this feature is entirely influenced by regression testing. The results show that the proposed method achieves over 81% fault localization in the Top-3 rankings. It also improvesEXAM score, with over 68% of programs scoring 10 or less via test case selection and 63% via test case reduction. Compared to spectrum-based techniques, the proposed method located 54% more faults than Tarantula and Ochiai, and 45% more thanJaccard in theTop-3 rankings. Additionally, it outperforms existing methods in EXAM score while using only 40% to 80% of the test suite. These findings highlight the importance of incorporating historical data, code structure, runtime behavior, and test case to improve fault localization efficiency in evolving software.