A Novel Way of Assessing Software Bug Severity Using Dictionary of Critical Terms

Abstract

Due to increase in demands of software and decreased delivery span of software, assuring the quality of software is becoming a challenge. However, no software can claim to be error free due to the complexity of software and inadequate testing. There is a well-known principle of testing, which states that exhaustive testing is impossible. Hence, maintenance activities are required to ensure smooth functioning of the software. Many open source software provides bug tracking systems to aid corrective maintenance task. These bug tracking systems allow users to report the bugs that are encountered while operating the software. However, in software maintenance, severity prediction has gained much attention recently. Bugs having higher severity should be fixed prior to the bugs having lesser severity. Triager analyzes the bug reports and assesses the severity based upon his/her knowledge and experience. But due to the presence of a large number of bug reports, it becomes a tedious job to manually assign severity. Thus, there is growing need for making the whole process of severity prediction automatic. The paper presents an approach of creating a dictionary of critical terms specifying severity using two different feature selection methods, namely- info gain and Chi square and classification of bug reports are performed using Naïve Bayes Multinomial (NBM) and K-nearest neighbor (KNN) algorithms.