The Bug Prediction Model Using Mutation Metrics

Please enable javascript in your browser.

Mohebbi, Ali | 2018

774 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 51305 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Mirian Hosseinabadi, Hassan
Abstract:
Software developers notice existence of faults by report of a fault in issue tracking systems or a failure in software tests. Then they try to locate the bug and understand the problem. Early detection of fault results in saving time and money and facilitates debugging process. Prediction models can be built and used easily by modern statistical tools. Software metrics are the most important part of prediction models. Therefore higher performance in models can be achieved using new and effective metrics. In this study, process metrics and metrics that built base on mutation analysis used and resulting models evaluated. In addition to using process metrics with mutation metrics, two group of metrics named mutation base process metrics and mutation-process hybrid introduced for building prediction models. Results showed that mutation metrics can improve prediction performance of process metrics. Although mutation based process metrics have a predictive value, they can not perform better than mutation metrics. Also mutation-process hybrid metrics can improve performance in prediction models significantly
Keywords:
Software Testing ; Defect Prediction ; Mutation Metries ; Process Metrics