Multiple-Bug Oriented Fault Localization: A Parameter-Based Combination Approach

Abstract

To improve the efficiency of localization technique, a parameter combination approach to direct multiple faults localization is proposed in this paper. We consider bisection methods as clustering rules to partition failed test cases into smaller fault-focused clusters with different kinds of combinations. Correlation coefficient can be a tolerance to accept or reject that one cluster aims at specific faults. A statistical approach, a cross tab-based technique will be adopted to help find each bug in a code block soon after. An efficiency comparison between parameter-based technique and Tarantula (using one-bug-at-a-time strategy) on the Siemens Suite will be carried out. The result in this paper implies that more bugs a program contains, more efficient Parameter-Based Combination technique (hereafter referred to as PBC) is. That means PBC has a better performance in multiple faults localization field.