EMBFL: Ensemble of Mutation based techniques for effective Fault Localization

Abstract

Finding locations of faults in a program is a crucial activity in reliable and effective software development. A large number of fault localization techniques exist, however, none of these techniques outperforms all other techniques in all,circumstances for all kinds of faults. Under different circumstances, different fault,localization techniques yield different results. In this study, we have proposed Ensemble of Mutation Based techniques for effective Fault Localization (EMBFL). EMBFL classifies statements of a program into Suspicious and Non-Suspicious sets. The model we have used in our research is straightforward and intuitive because it is based solely on information regarding statement coverage and test case execution results. This helps to reduce the search space significantly. Our proposed EMBFL approach, on average, is 31.34% more effective than the techniques for fault localization that currently exist such as DStar (D*), Tarantula, Back Propagation Neural Network, etc.