International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056 Volume: 03 Issue: 11 | Nov -2016 www.irjet.net p-ISSN: 2395-0072 © 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page 1202 An Effective Approach for Automatic Bug Triage Techniques Sonali Thaokar 1 , Prof. Pragati Patil 2 1 Sonali Thaokar, Dept. of Compuiter Science & Engineering, AGPCE, Maharashtra, India 2 Prof. Pragati Patil, Dept. of Compuiter Science & Engineering, AGPCE, Maharashtra, India ---------------------------------------------------------------------***--------------------------------------------------------------------- Abstract - The process of fixing a bug is called bug triage that goal is assigned to a developer for new coming bug. In a software firm, they spend their time and price to manage the bugs. So to reduce time and price of manual work in software firm they use automatic bug triage. By automatic bug triage, find predicted developer to resolve the bugs. In proposed approach, we used data reduction techniques and machine learning algorithm. To enhance standard of data, we used data reduction techniques, for that feature selection and instance selection techniques are used. We used feature selection and instance selection techniques at the same time to improve the accuracy of automatic bug triage. Also, we used machine learning technique for bug triaging system. We have added a new module here which will describe the status of the bug like whether it assigned to any developer or not and it is rectified or not. In addition, the load between developers based on their experience is re-balanced. The experimental result shows high prediction accuracy by using data reduction techniques and machine learning algorithm. Key Words: Bug data reduction, Feature selection technique, Instance selection technique, Machine learning algorithm technique, Bug Triage. 1.INTRODUCTION In a software firm, bug fixing is very time consuming process. Many open source software projects have an open bug repository that makes it possible for each developer and users to publish defects or issues in the software, suggest possible enhancements, and remark on existing bug reports. In large open source software project have the bug repository that store the details of the bug. For large open source software project, the quantity of every day bugs is so substantial which makes the triaging process more challenging and difficult.There are two challenges associated with bug data that will have an effect on use of bug repositories in software development tasks, specifically the large scale data and low quality data. In a bug repository, a bug is kept up as a bug report, which record in the form of text that reproducing the bug and update as per the status of bug fixing. Manual bug triage is very time consuming for software firm because they spend their time and cost to manages the bug. The process of assigning a proper developer for fixing bug is the bug triage. By using automatic bug triage, software firm manages the bug easily and it save the time and cost of manual work. For automatic bug triage we used machine learning proposed data reduction techniques i.e. feature selection and instance selection techniques. By using these techniques reduce the bug data to save the labor price of developer and enhance the quality of bug data and increase the accuracy of bug triage. Section [2] describes background and Section [3] describes the system architecture of the proposed system. Section [4] describes the data set collection. The details of instance selection, feature selection is given in Section [5] implementation, the snapshot of proposed system given in Section [6] and concluded in Section [7]. 2. BACKGROUND Xuan et al. [1] proposed to reduce the bug data used instance selection and feature selection techniques. Their approach effectively reduced the data scale by using data reduction techniques improved accuracy of bug triage. Anvik et al. [2] they used supervised learning machine algorithm to assignment of bug report to the potential developer. They reached precision levels of 57% and 64% on Eclipse and Firefox respectively. Alenezi et al. [3] in this approach used term selection method to recognize the good quality of bug report and improve accuracy and used naïve bayes classifier to predict the developer for each new bug report. They result shows that improved F-score. Anjali et al. [4] proposed Domain Mapping Matrix to predicting the best suited developer to resolve the newly bug reports. They achieved an efficiency of 86% for top-10 and 97% for top-20 developer ranking list. Cubranic et al. [9] this approach used text categorization dominates the existing bug triage. The first work of bug triage is a supervised text categorization approach using Naive Bayes. Their approach achieved 30% accuracy. Nhan Minh Phuc [15] proposed To automatically detect duplicate bug reports, used Class-Feature-Centroid (CFC). The recall rate is improved by 10% for 20 predictions for SVN and AgroUML.