Data reduction for bug triage using effective prediction of reduction order techniques

Abstract

Translate Article

A large open source project consists of a wide range of bug reports. In open source project, bug reports are available and these reports can be modified by anyone. Bugs are software defects whose prediction is highly difficult. To detect the bugs, machine learning classifier has been proposed. It segregates the bug reports and developers and it learns the type of report suitable for each developer. Bug triage is the process of assigning the bug for the appropriate developer. The techniques include preprocessing, machine learning classifier, instance selection and feature selection. The aim of this paper is to attain a data set reduction in bug triage by including the representative values along with the statistical values of the bug data set. Our work considers the dataset from the open source project Eclipse. We focus on reducing the data scale and thereby improving the accuracy. This can be achieved by building a representative model for prediction of reduction orders by including the summary, metadata. Our proposed work attains an accuracy result of 96.5% that is better when compared with existing work.