Deep-Learning Approach with DeepXplore for Software Defect Severity Level Prediction

Abstract

Fixing the defects of earlier releases and working on fast and efficient fixing of those software defects is detrimental for the release of further versions. Bug tracking systems like Bugzilla get thousands of software defect reports every day. Manually handling those report to assign severity to the defects is not feasible. Earlier traditional Machine Learning methods have been used to predict the severity level from the defect description. This paper presents different deep learning models to predict defect severity level. Furthermore, the deep neural network was tested using a framework developed similar to that DeepXplore. Different word-embedding techniques, feature-selection techniques, sampling techniques and deep learning models are analyzed and compared for this study. In this paper, we have considered Descriptive statistics, Box-plot, and Significant tests to compare the developed models for defect severity level prediction. The three performance metrics used for testing the models are AUC, Accuracy and Neuron Coverage. This is a preliminary study on DNN testing on this dataset. Thus, the paper focuses on DeepXplore DNN testing technique. However further studies would be undertaken on comparative analysis of different DNN testing techniques on this dataset.