A Strategy to Evaluate the Fault-focused Clustering Performance of Distance Metrics and SBFL Formulas in Parallel Fault Localization
Yihao Li, Pan Liu, Xiao Qiang Zhao, Xiaobin Sun, Yongtao Li · 2021
Parallel fault localization (PBL) is a common practice to locate multiple bugs at the same time in spectrum-based fault localization (SBFL). To do so one usually needs to group failed test cases that are likely due to the same bug into the same fault-focused clusters. To evaluate the fault-focused clustering performance, two impact factors distance metric and SBFL formulas are critical to the performance of fault-focused clustering as well as the performance of PBL. This paper proposes a strategy to evaluate the fault-focused clustering performance of distance metrics and SBFL formulas from the perspective of perfect clustering (PC) where the ideal fault-focused clusters are already known.