![]() | Only 14 pages are availabe for public view |
Abstract Defective software modules are causing software failures, increasing development and maintenance costs, and decreasing customer satisfaction. However, understanding the impact of defects in various business applications is an essential way to improve software quality. In software companies, bugs repositories become precious sources for solving issues in different projects. There are lots of factors related to bugs fixing that include identifying the defect severity and its category. Severity which is defined as the impact of the bug in the software system. Identifying bug severity is an important task commonly used to avoid misclassification of the bug. While specifying bugs category before debugging will lead to better handling of selecting the proper destination to fix it. For instance, some reporters usually raise issues about the severity as critical for rapid repairing which is not a correct behavior. This is from one side, from the other; the study would expect that functionality bugs are fixed faster than other types of bugs due to their critical nature. However, prior researches have often treated all bugs as similar which is not a good way. The presented thesis studies the feasibility of using efficient data mining techniques to establish a model to predict the severity and category of a bug given the summary of the bug only. Specifically, this can be done by using open source repositories of the Eclipse and GNOM repositories for modeling and predicting bug severities. The study compares the proposed algorithm and some other bench-marked algorithms. By ii |