Research On Software Bug Report Localization Based On Learning To Rank And Information Expansion

With the rapid growth of software project requirements,there will inevitably be some bugs in the process of software design and implementation.When software project users and testers find that there are situations that are different from the expected results during program operation,they will promptly report the bugs they found to the bug tracking system.But for large open source software projects,the bug tracking system receives a large number of bug reports every day.It is extremely time-consuming and energy-consuming to rely on developers to manually check and find the location of the relevant source code.Therefore,it is very important to automatically locate the bugs in the bug report to the relevant source code.The bug report location method based on information retrieval has always been a research hotspot in this field because of its fast speed and low resource consumption.But it also has shortcomings.The method mainly relies on the textual similarity between the bug report and the source code file,and there is a vocabulary gap between the bug report written in natural language and the source code written in programming language,which makes the method based on information retrieval Unable to take advantage.Especially the method-level bug report location research,because the method body has less code and contains less text information,which makes the method-level bug report location accuracy unable to be improved.In response to these problems,this paper studies the method-level and file-level bug report location methods respectively.The specific work is as follows:For the method-level bug report location research,firstly,from the three aspects of semantic similarity,time proximity,and call dependence,expand the text information of the method body,enrich its semantic information,and make up for the lack of text information in the method body.The bug report is the problem of insufficient location accuracy.In addition,the metadata feature of the software project is combined to improve the location accuracy.Finally,the linear fusion of these features is used as a scoring function,and the methods that may have bugs are recommended to the developers according to the relevance scores from large to small.For the file-level bug report location research,first build a developer social network through reviewers under historical similar bug reports,use 7 metrics to measure the importance of each developer,and find the most suitable developer to repair the bug,then the source code files repaired by these developers are likely to still be bug files.Mining the same metadata feature extraction and feature fusion combination as the method-level bug report location research.And recommend possible source code files to developers.