A Data Mining Model to Predict Software Bug Complexity Using Bug Estimation and Clustering

Software defect(bug) repositories are great source of knowledge. Data mining can be applied on these repositories to explore useful interesting patterns. Complexity of a bug helps the development team to plan future software build and releases. In this paper a prediction model is proposed to predict the bug's complexity. The proposed technique is a three step method. In the first step, fix duration for all the bugs stored in bug repository is calculated and complexity clusters are created based on the calculated bug fix duration. In second step, bug for which complexity is required its estimated fix time is calculated using bug estimation techniques. And in the third step based on the estimated fix time of bug it is mapped to a complexity cluster, which defines the complexity of the bug. The proposed model is implemented using open source technologies and is explained with the help of illustrative example.