Doctor of Philosophy
In this thesis we developed a new fault localization process to localize faults in object oriented software. The process is built upon the "Encapsulation'' principle and aims to locate state-dependent discrepancies in the software's behavior. We experimented with the proposed process on 50 seeded faults in 8 subject programs, and were able to locate the faulty class in 100% of the cases when objects with constant states were taken into consideration, while we missed 24% percent of the faults when these objects were not considered. We also developed a customized data mining technique "Associated sequence mining'' to be used in the localization process; experiments showed that it only provided slight enhancement to the result of the process. The customization provided at least 17% enhancement in the time performance and it is generic enough to be applicable in other domains. In addition to that we have developed an extensive taxonomy for object-oriented software faults based on UML models. We used the taxonomy to make decisions regarding the localization process. It provides an aid for understanding the nature of software faults, and will help enhance the different tasks related to software quality assurance. The main contributions of the thesis were based on preliminary experimentation on the usability of the classification algorithms implemented in WEKA in software fault localization, which resulted in the conclusion that both the fault type and the mechanism implemented in the analysis algorithm were significant to affect the results of the localization.
Ali, Shaimaa, "Localizing State-Dependent Faults Using Associated Sequence Mining" (2013). Electronic Thesis and Dissertation Repository. 1388.