Improving the quality of Software Fault localization using Rank Aggregation

Fault localization is one of the tasks in the software debugging process, it aims to identify the position of faults in the software system under test. Spectrum based fault localization techniques are a lightweight and efficient sub-branch of this field. These techniques utilize executive behavior of the program under test and assign suspiciousness scores to program entities using risk evaluation formulas. Many studies proposed a risk evaluation formula to localize the faults in the software system, but they don’t have the same quality for all cases. In this paper, we proposed a method to improve the performance of fault localization in most cases. We aggregate results of spectrum-based fault localization techniques with the MEDRank algorithm and identify the location of faults more accurate in various cases. For the aggregating process, we select some risk evaluation formulas and use the rank aggregation algorithm to consensus the results. We evaluate our method with individual spectrum-based fault localization techniques and another rank list aggregation method. Our method can improve the fault absolute rank with an effectiveness rate of 34.6% on Siemens benchmark but over considered methods, the effectiveness of the best is 39.1%. The results show that the proposed method has better quality to locate fault compared with other techniques.