Software Defect Classification using Bayesian Classification Techniques

Classifying a defect is an important activity for improving software quality. It is important to classify defects as they contain information regarding the quality of processes and product. The information gathered from defects can be used to track the future projects, and to improve its processes. Considering the need to classify the defect and to gain insight knowledge of defect details, this paper attempts to analyze software defect using Bayes net and Naïve Bayes Classification techniques. It is very difficult to produce defect free software product, however, the main purpose of any software engineering activity is to prevent defects from being introduced in the first place. As fixing of software defects are expensive and time consuming, various defect prediction techniques and defect tracing mechanism are being used to prevent software defects from occurring. The aim of this paper is to show the comparative analysis of software defect classification using Bayes net and Naïve Bayes classification techniques in terms of accuracy, Precision, Recall etc. The Naive Bayesian classifier assumes that all variables contribute toward classification and that they are mutually independent. A Naive Bayesian model leads to a simple prediction framework that gives good result which makes it particularly useful for very large datasets like public NASA MDP repository and the same is experimented in this study. The study revealed that, the performance of Naïve Bayes classification for software defect outperforms Bayes net classification method