Predicting Bug severity using Classification on Clustered Bugs Data.

Bug Triaging is an inevitable step in software testing process. It assigns a potential developer to resolve the bug and if it is done manually, it takes lot of time and resources. Hence a automatic bug triage system is developed to address the issue. In addition to predicting the developer, it is important to understand how quickly the bugs are addressed based on severity factor. Bug severity prediction based on classification algorithms in data mining helps to predict the severity of the bug. The system also studies the effect of clustering before classification. Here we group similar bugs to a cluster based on the problem title attribute. Thus classification was then applied to the clusters obtained, to assign severity labels to bugs based on priority, product and component attributes. A comparative study of different combinations of classification and clustering algorithm on the performance of prediction is also undertaken in this work.