一种针对软件缺陷自动分派的开发者推荐方法  

Developer Recommendation Method for Automatic Software Bug Triage

作　　者：刘海洋 马于涛[1] 

机构地区：[1]武汉大学软件工程国家重点实验室,武汉430072

出　　处：《小型微型计算机系统》2017年第12期2747-2753,共7页Journal of Chinese Computer Systems

基　　金：国家"九七三"重点基础研究发展计划项目(2014CB340404)资助;国家自然科学基金项目(61272111)资助

摘　　要：开源软件的缺陷管理是其软件质量保障的一种重要手段,而缺陷的高效分派是大型开源软件缺陷管理的一个棘手问题.为了提高缺陷分派的效率,本文提出了一种简单易用的针对软件缺陷自动分派的开发者推荐方法,其核心思想是利用LDA主题模型(刻画开发者技能)、开发者合作网络(刻画开发者之间的合作关系)构造(内容+关系)混合策略.针对大型开源软件项目Eclipse和Mozilla的十万级已修复缺陷的实验表明,在选取合适的参数和分派策略情况下,本文所提方法的开发者推荐的准确率分别达到了46.7%和33.4%,比基准的LDA+k NN方法的推荐准确率分别提高了209.3%和131.9%,从而验证了其有效性.Software defect/bug management has become an important means of quality assurance for open source software （ OSS ）, and efficient bug triage is a challenge to bug management in large-scale OSS projects. To improve the efficiency of bug triage, in this paper we present a simple, easy-to-use developer recommendation method for automatic software bug triage. In particular, its core is a hybrid strategy built by using LDA topic modeling （for describing developer＇s expertise ） and a developer collaboration network （ for describing the collaboration among developers ） based on bug tossing. We conducted an experiment on hundreds of thousands of bugs in the Eclipse and Mozilla projects,and found that when selecting appropriate parameters and tossing strategy, the accuracies of our method for developer recommendation in Eclipse and Mozilla achieved 46. 7% and 33.4%, respectively. Moreover, the accuracies of our method for developer recommendation in Eclipse and Mozilla were increased by 209.3% and 131.9%, respectively, compared with the baseline LDA ＋ kNN method, which indicates the effectiveness of our method.

关 键 词：缺陷分派 开源社区 开发者合作网络 主题模型 混合策略 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]