Determining Bug Severity using Machine Learning Techniques

Abstract

Software Bug reporting is an integral part of software development process. Once the Bug is reported on Bug Tracking System, their attributes are analyzed and subsequently assigned to various fixers for their resolution. During the last two decades Machine-Learning Techniques (MLT) has been used to create self-improving software. Supervised machine learning technique is widely used for prediction of patterns in various applications but, we have found very few for software repositories. Bug severity, an attribute of a software bug report is the degree of impact that a defect has on the development or operation of a component or system. Bug severity can be classified into different levels based on their impact on the system. In this paper, an attempt has been made to demonstrate the applicability of machine learning algorithms namely Naïve Bayes, k-Nearest Neighbor, Naïve Bayes Multinomial, Support Vector Machine, J48 and RIPPER in determining the class of bug severity of bug report data of NASA from PROMISE repository. The applicability of algorithm in determining the various levels of bug severity for bug repositories has been validated using various performance measures by applying 5-fold cross validation.