Bug localization in software using NSGA-II

摘要

Finding bugs in a software is a cumbersome and tedious task. When a new bug is reported, the developers find it challenging to replicate the unexpected behavior of the software, in order to fix the original fault. In this paper, an automated model is presented to find and sort the classes present in the source code according to their proneness of containing a bug, depending upon the bug reports. The model uses a text mining approach and recommends a list of classes based upon the lexical similarity between bug reports and the API descriptions, and also the changes previously recommended during bug fixing. To maximize the similarity index and at the same time reduce the number of classes recommended, a Non-dominant Sorting Genetic Algorithm (NSGA-II) is employed. This model was evaluated on three java based open source applications and it is observed that the model created using multi-objective NSGA-II outperforms the traditional methods of bug localization.