计算机科学与探索 ›› 2017, Vol. 11 ›› Issue (5): 700-707.DOI: 10.3778/j.issn.1673-9418.1609033

• 学术研究 • 上一篇    下一篇

深度学习在缺陷修复者推荐中的应用

胡  星,王千祥+   

  1. 北京大学 高可信软件技术教育部重点实验室,北京 100871
  • 出版日期:2017-05-01 发布日期:2017-05-04

Application of Deep Learning in Recommendation of Bug Reports Assignment

HU Xing, WANG Qianxiang+   

  1. Key Lab of High Confidence Software Technologies (Peking University), Ministry of Education, Beijing 100871, China
  • Online:2017-05-01 Published:2017-05-04

摘要: 目前许多软件项目使用缺陷追踪系统来自动化管理用户或者开发人员提交的缺陷报告。随着缺陷报告和开发人员数量的增长,如何快速将缺陷报告分配给合适的缺陷修复者正在成为缺陷快速解决的一个重要问题。分别使用长短期记忆模型和卷积神经网络两种深度学习方法来构建缺陷修复者推荐模型。该模型能够有效地学习缺陷报告的特征,并且根据该特征推荐合适的修复者。通过与传统机器学习方法(如贝叶斯方法和支持向量机方法)进行对比,该方法可以比较有效地在众多开发者中找出合适的缺陷修复者。

关键词: 缺陷追踪, 缺陷报告分配, 深度学习

Abstract:  Open source projects typically support an open bug repository to which developers and users can report bugs. As the increase in bug reports and developers, it is a challenge to assign large amounts of bug reports effectively to the appropriate developers. This paper applies two deep learning approaches, long-short term memory and convolutional neural network, to learn the features of bug reports and then makes assignments. Deep learning approaches are expert in learning features and making assignments effectively with the help of features. Compared to the traditional machine learning approaches such as Bayesian learning and support vector machine, the proposed approach can assign bug reports to developers effectively.

Key words: issue tracking, bug report assignment, deep learning