The Research On The Fixer Recommendation Method Of Software Bugs

Software bug fixer recommendation is the process of recommending the appropriate fixer to bug triager to resolve bugs in bug assignment.Conventional methods assign bug reports manually.However,along with the increasing of functional requirements and software scale,the number of software bugs have increased dramatically,especially in the open source software domain.As such,manually assigning bug reports is an extraordinarily time-consuming and error-prone task.Therefore,it is essential to propose an effective method of recommending software bug fixer.Machine learning is very effective in recommending a fixer.This thesis studied a method of using the classical Support Vector Machine(SVM)to recommend bug fixer for open source software.On the other hand,it is essential to exploit a new source by considering that most of the existing bug fixer recommendation methods are overdependent on the prior fixing data of fixers and are often underperform when lack of prior fixing data.Well developed Community Question Answering platform(CQA),e.g.,Stack Overflow,offers opportunities for fixers to communicate when they meet technical issues.Hence,extensive information in those CQA reveals expertise of fixers,which holds excellent latent value for fixer recommendation.Based on aforementioned,thesis main contributions are as follows:1.Based on the analysis of the research status of different machine learning methods in the field of bug fixer recommendation,this thesis focused on utilizing SVM to address bug fixer recommendation problem in open source software filed.Taking 6654 bug reports from GitHub as experimental data,this thesis investigated the data preprocessing and training steps of SVM in these open source data in detail.Through the experiment,this thesis analysed the effectiveness and drawbacks of this method in the field of fixer recommendation.2.The thesis designed a method of recommending bug fixer based on CQA information.It consists of two sections: the initial section used tags in CQA information as a bridge associating bug reports and the CQA information,then set different weights to corresponded tags and combined the number of upvotes from CQA information to measure the expertise of fixers.The other section based on the fluctuation of fixers' expertise that expertise of the fixers shifts over time as they work on different projects to estimate the time-aware of fixers' fixed work.Finally,this thesis weightd the output of two above sections as final results to recommend fixers.Using CQA information from Stack Overflow and bug reports from GitHub as experimental data,the experimental results showed that the method designed in this thesis increased 0.37%~19.34% in Top5(%)recommendation accuracy compared with the existing fixer recommendation methods.