Exploration of Machine Learning Techniques for Defect Classification

Abstract

To develop good quality software product, there is a need of continuous defect identification and classification in each module before delivering a software product to the customer. Developing software needs proper managing of the available software resources. To deliver a software product on time, developing quality software products, Information Technology (IT) industries normally use software tools for defect detection. Based on severity, defects are detected and classified. This can be automated to reduce the development time and cost. Nowadays, machine learning algorithms have been applied by many researchers to accurately classify the defects. In this paper, a novel software defect detection and classification method is proposed and neural network models such as Probabilistic Neural Network model (PNN) and Generalized Regression Neural Network (GRNN) are integrated to identify, classify the defects from large software repository. Defects are classified into three layers based on the severity in the proposed method abstraction layer, core layer, and application layer. The performance accuracy of the proposed model is compared with MLP and J48 classifiers.