Research On Fault Localization Using Conditional Probability Formulas

With the rapid development of modern science and technology, computer software has been widely used in every field of people’s life. So people pay more attention to its quality. Software testing plays an important role in the software development process. And fault localization is an important step of software testing in removing the software defects. So improving the efficiency of fault localization has important significance.As the continuous development of software testing techniques, automated test-ing is gaining its popularity in practical testing. It can help the programmers locate faults. Automation testing can improve the efficiency of testing compared to tradition-al manual testing. Among the automation testing techniques, fault localization based on spectrum only needs the source code of testing programs and the set of test cases. Programmers can achieve a good efficiency by spectrum-based fault localization.This thesis analysis spectrum-based fault localization, and propose our method: fault localization using conditional probability formulas. We attend to improve the efficiency of fault localization by a comprehensive consideration of the two aspects which affect the location of the fault:executed source code and fault test cases. In order to assess the effectiveness of the method we propose, we also conduct a set of empirical studies.The main contributions of this thesis can be summarized as follows:1. Topics on fault localization in software testing are reviewed.First, the ba-sic concepts and classification of fault localization are introduced. Then, some concepts and definitions for fault localization are discussed. Finally, related work about fault localization are classified into five categories:spectrum-based, dependency-based, state-based, method-based and test case reduction based. The main ideas of these approaches are presented respectively especially the spectrum-based technology.2. A fault localization approach based on conditional probability formulas, is proposed. First, the motivation of the method is introduced through an example. Then the framework of the method is described. Finally, a new fault localization approach based on conditional probability formulas is proposed according to the the framework. Specific design and implementation of the method are described.3. An empirical study on the effectiveness of the fault localization method based on conditional probability formulas is conducted. The fault localization method based on conditional probability formulas is implemented on 8 experimental sub-jects. And comparing it to Tarantula and Ochiai using the common evaluation metric. Finally, we analysis the results of the experiments to verify the effective-ness of the method proposed in this paper.