基于多目标优化算法NSGA-II推荐相似缺陷报告  

Recommending Similar Bug Reports Based on Multi-Targets Optimization Algorithm NSGA-II

作　　者：樊田田 许蕾[1,2] 陈林[1,2] FAN Tian-Tian;XU Lei;CHEN Lin(Department of Computer Science and Technology,Nanjing University,Nanjing 210023;State Key Laboratory for Novel Software Technology,Nanjing University,Nanjing 210023)

机构地区：[1]南京大学计算机科学与技术系,南京210023 [2]南京大学计算机软件新技术国家重点实验室,南京210023

出　　处：《计算机学报》2019年第10期2175-2189,共15页Chinese Journal of Computers

基　　金：国家“九七三”重点基础研究发展规划项目（2014CB340702）;国家自然科学基金（61272080,91418202,61403187）资助~~

摘　　要：在软件开发过程中,开发人员会收到用户提交的大量缺陷报告.若修复缺陷报告中问题涉及到的相同源代码文件数目超过一半,则称这些缺陷报告为相似缺陷报告.给开发人员推荐相似缺陷报告能够有效节约开发人员修复缺陷的时间.该文提出一种基于多目标优化算法NSGA-Ⅱ推荐相似缺陷报告的方法,即在推荐尽可能少的相似缺陷报告情况下,使得缺陷报告间的相似度尽可能大.为此,利用缺陷报告的摘要和描述信息,该文采用TFIDF和Word Embedding两种方法,从历史缺陷报告中找出相似的缺陷报告,并采用基于搜索的多目标优化算法NSGA-Ⅱ来保证推荐的相似缺陷报告数目尽可能少.实验数据集是6个开源项目(AspectJ、Birt、Eclipse UI、JDT、SWT和Tomcat).与采用单目标算法相比,该文方法在推荐相似缺陷报告的准确率、平均准确率均值、平均序位倒数均值都有提高,其中,在Top@1准确率、平均准确率均值、平均序位倒数均值上分别比Yang方法提高125.5%、67.7%和62.75%.During the process of software development,developers usually receive a large number of bug reports,submitted by users,and they must handle these bug reports.Bug report is a description of the information about the problems encountered by software users.Developers fix bugs according to bug reports.Among them,if the fixing issues in bug reports are concerned with the same source code files(more than half),then these bug reports are called as the similar ones.Recommending similar bug reports for developers is quite important,since it can save a lot of time and improve the efficiency.At present,the existing methods usually recommend similar bug reports based on information retrieval methods,and handle the title and description information of bug reports.Considering the product and component information in the bug report,the similarity between the query bug report and the historical bug report is calculated,and then the recommendation list is returned for developer’s reference.And the correlation between the previous and current bug reports is not necessarily typed.This paper proposes a recommendation method for similar bug reports based on a multi-targets optimization algorithm(NSGA-Ⅱ),and the number of recommendations is as small as possible,with the maximum similarities among bug reports.NSGA-Ⅱalgorithm is a genetic algorithm,based on multi-objective optimization,which is used to solve the shortages of existing methods.The main challenges of using genetic algorithm to solve problems in recommending similar bug reports are:(1)transferring the problem of recommending similar bug reports as the problem using genetic algorithm,which need to consider how to represent individuals and how to form a population;(2)selecting appropriate fitness function to evaluate the quality of individuals in the population,since the fitness function is greater,the individual is better,and the probability of being inherited to the next generation is greater.In order to avoid local optimization,we need to select suitable selection op

关 键 词：相似缺陷报告推荐 多目标优化 空间向量模型 词嵌入模型 NSGA-Ⅱ算法 软件工程 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]