Software bug localization using Pachinko Allocation Model

摘要

Bug localization is the process of identifying the elements of source code that require modification to fix the bug. By automating the task of bug localization efficiently, the cost of software can also be reduced. For performing bug localization, many Information Retrieval models have been used in past. In this paper, bug localization has been performed using Pachinko Allocation Model (PAM). PAM is also an IR model, which falls under the category of topic models and has not been used for locating bugs yet. This paper describes the proposed PAM based approach for bug localization at file level. The PAM based approach is compared with LDA based approach and it has been shown that PAM based bug localization performs better as compared to LDA based bug localization. For evaluating the performance of PAM and LDA based approaches, the datasets downloaded from two open source projects, i.e. Rhino and ModeShape, have been used.