Syntactic and semantic feature extraction and preprocessing to reduce noise in bug classification

Abstract

In software industry a lot of effort is spent in analyzing the bug report to classify the bugs. This Classification helps in assigning the bugs to the specific team for Bug Fixing according to the nature of the bug. In this paper, we have proposed a data mining technique applying syntactic and semantic Feature Extraction to assist developers in bug Classification. Extracted features are organized into different feature groups then a specific preprocessing technique is applied to each feature group. The applied methods have reduced the noise in the bug data compared to traditional approach of word frequency for text categorization. We have analyzed our approach on a collection of bug reports collected from a networking based organization (CISCO).The experiments are performed using Naive Bayes Multinomial Model and Support Vector Machine on features obtained after preprocessing. � 2012 Springer-Verlag.