Downloads: 108 | Views: 220

Case Studies | Computer Science & Engineering | India | Volume 5 Issue 10, October 2016

Improving Auto Bug Triage by Effective Data Reduction

Roshna V. Sangle | Rajendra D. Gawali

Abstract: Large open source software projects like Mozilla, Firefox, Eclipse etc. receive huge number of submitted bug reports daily basis. Manual Triaging of these upcoming reports is error-prone and time consuming process. The purpose of bug triaging is to assign potentially experienced developers to upcoming bug reports. Thus to reduce cost and speedup bug triaging, this paper presents an automatic approach to predict a developer with approximate time required to solve the upcoming report. In proposed system data set reduction is achieved through techniques like stemming, stop word removal, Instance selection and Feature selection on bug data set, which improve the scale and quality of bug data. The simultaneous usage of instance selection and feature selection reduces the scales on bug dimension and word dimension which improves the accuracy of bug triage. The combination of feature selection algorithm, statistics (CHI2) and instance selection algorithm, Iterative Case Filter (ICF) is applied in proposed paper. Then Naive Bayesian classifier is used to predict the expert developer to fix the upcoming bug. This paper also focuses on how to assign any upcoming bug to new developer whose bug fixing history is not available in training dataset.

Keywords: Bug Triage, Stemming, Stopwords, Instance selection, Feature Selection, Training dataset, cold-soft etc

Edition: Volume 5 Issue 10, October 2016,

Pages: 1987 - 1990