![]() | Only 14 pages are availabe for public view |
Abstract Recent software systems usually feature an automated failure reporting system, with which a huge number of failing traces are collected every day. In order to prioritize fault diagnosis, failing traces due to the same fault are expected to be grouped together. Recommendation of fault location techniques have taken a lot of interest in the last decade. However. these techniques tend to present a ranked list of bug predictors ivhic:h is not convenient for software containing multiple bugs. This thesis proposes an enhanced bi-clustering algorithm for automatic multiple soft¬ware bug isolation. Proposed algorithm avoids the drawback of current algorithm that is putting execution trace of the same bug in more than one bi-cluster. It also makes a step towards shrinking predictors list in order to be useful and not distracting to the developer. The algorithm is validated both quantitatively and qualitatively using standard data sets of feedback reports. |