Research On Software Fault Location Method Based On Cross Matrix Statistics

Fault is one of the important risk factors to the quality of software production,and the severity is increasing.So how to locate faults accurately as soon as possible when the software failure has occurs,has become a growing topic.Nowly the fault location mainly rely on developers manual debugging,fault positioning efficiency depend on the programmer’s technical level and development experience,so it is a time-consuming process.Automatic fault location technique can be combined with the automatic test technology,it has important practical significance for improving the efficiency of software fault localization.Automatic fault location method includes fault location technology based on static analysis(SABFL)and the fault location technology based on the analysis of the test(DTBFL).Based on the fault location execution trace(TBFL)in recent years in the field of dynamic test is an active research topic.In TBFL’s research needs a large number of execution sequences of trajectories(success or failure)as the material,and these data surely existing a lot of execution trace sequence redundancy will greatly reduce the efficiency and accuracy of fault location.To solve this problem,This paper studies an improvement of software fault localization technology based on crossover matrix statistical method.This method firstly makes clustering reduction for all traces(successful or failure)before fault localization to eliminate the redundancy execution traces,and then stores the remaining execution traces to crossover matrix,and finally calculates the dubiety of the statements in the crossover matrix by the Crosstab algorithm and ranks them to get the fault report.Based on Siemens Suites having performed the experiment of the performance comparison before and after the traces clustering reduction,the experimental result verifies the effectiveness of the proposed method.