An effective fault localization approach based on PageRank and mutation analysis

Abstract

Mutation-based fault localization (MBFL) is a popular method based on mutation testing. MBFL applies a variety of operators to generate mutants and calculates the statement’s suspiciousness by counting the execution results of the test cases on the mutants. However, the tie problem of MBFL creates obstacles to accurate fault localization. The tie problem refers to that many statements have the same suspiciousness. To solve the tie problem, we propose a fault localization approach based on the PageRank algorithm and mutation analysis (PRMA). We first apply the PageRank algorithm to calculate the faultiness scores of the statements. Then, we weight the suspicious value of the statements with faultiness scores to solve the tie problem. Finally, the weighted suspicious values are sorted in descending order to generate a list, which is provided to developers for fault localization. To evaluate our approach, we conduct experiments on the real fault benchmark Defects4J and the artificial fault dataset Siemens. We compare PRMA with the traditional MBFL techniques (Metallaxis and MUSE) and recently proposed MBFL methods (MCBFL-hybrid-avg, SMFL and SMBFL). The experimental results show that our approach outperforms above comparison methods and improves the effect of fault localization in both quantity and accuracy.