An Automated Approach for Software Bug Classification

Abstract

Open source projects for example Eclipse and Fire fox have open source bug repositories. User reports bugs to these repositories. Users of these repositories are usually non-technical and cannot assign correct class to these bugs. Triaging of bugs, to developer, to fix them is a tedious and time consuming task. Developers are usually expert in particular areas. For example, few developers are expert in GUI and others are in java functionality. Assigning a particular bug to relevant developer could save time and would help to maintain the interest level of developers by assigning bugs according to their interest. However, assigning right bug to right developer is quite difficult for triager without knowing the actual class, the bug belongs to. In this research, we have classified the bugs in different labels on the basis of summary of the bug. Multinomial Naïve Bayes text classifier is used for classification purpose. For feature selection, Chi-Square and TFIDF algorithms were used. Using Naïve Bayes and Chi-square, we get average of 83 % accuracy.