Authors : S. Kirubakaran, K. Maheswari and K. Reshma Revathi
Abstract: Bug triage is an unescapable process in every software organization. A separate team in every software companies takes care of this process. The complete process occurs in manual which increases the production time and cost. One of the time taking tasks in bug triage process is assigning an appropriate developer to fix the new coming bugs than fixing that bug. Automation is the key solution for this problem. Since the bug reports are in free form of data, it is efficient to use data mining techniques to handle. In recent research, different techniques were applied by different authors trying to automate the bug triage process. But, up till now only 64% of triaging accuracy is achieved. Our survey shows that the decrease in accuracy is caused by one of the major problem called data reduction. In this study, we proposed a new framework called Auto-BugTriager which focuses on the elimination of data reduction problem to greater extent. Auto-BugTriager consist of three phases namely InfoZie, DataReduction and NBClassifier which works together to predict the recommendation list of expert developer for fixing the new bug. We have done the complete theoretical and experimental analysis of the proposed framework. Our analysis shows, that the proposed framework eliminated the problem of data reduction to greater extent, thus the data quality and accuracy of bug triage is increased.
S. Kirubakaran, K. Maheswari and K. Reshma Revathi, 2016. Auto-Bug Triager for Assisting Manual Bug Triage. Asian Journal of Information Technology, 15: 1334-1339.