Assessing the Severity of Software Bug Using Neural Network

Bug severity defines that to what extent the bug influence the software. Bug reports are misclassified affecting the performance of prediction and lead to biasness. Multitude relevant and irrelevant bugs are reported through bug-tracking systems, and it becomes necessary for the developer to give precedence to the bugs reported. The critical bugs need to be fixed immediately, and minor bugs can be fixed later on the basis of available resources. In this paper, we define different severity levels and assort them on the basis of their severity levels. It becomes problematic to assort tons of bugs manually, so we have used text mining approach to congregate words pointing severity of the reports obtained from Bugzilla for different components of ECLIPSE. The TF-IDF method is used to extract features and make dictionary of critical words. We have adopted neural network for the classification of bug reports and measured its performance.