大众点评驱动的开源软件演化机理和众包式改良方法研究

项目名称： 大众点评驱动的开源软件演化机理和众包式改良方法研究

项目编号： No.61272111

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 马于涛

作者单位： 武汉大学

项目金额： 80万元

中文摘要： 从大众用户在互联网上与开源社区不断交互过程中涌现出的群体智能，推动了开源软件OSS快速发展。评估和预测群体智能对OSS项目发展的影响，并加以利用来改良软件品质，是当前OSS基础研究面临的挑战性问题。研究旨在：探寻OSS演化与大众点评之间的内在联系，从机理上揭示用户点评形成的大众口碑和软件版本更新之间的相互影响和演变规律；构建大众点评驱动的众包式改良方法，从手段上提高缺陷分派、重现及传播危害评估的效率；开发原型工具并开展实证研究，从实践上提升OSS项目开发和管理水平。创新贡献为：用大众口碑代替传统的用户满意度，发现并阐明新动因驱动的OSS演化新特性和新机制，并利用时间序列分析方法创建OSS项目发展潜力预测模型；挖掘众包方式下更大开发者社区中的社群及其群体特长，构建基于马尔可夫链的缺陷再分配网络模型，应用协同过滤和随机游走推荐合适的开发者，比现有经典方法的正确率提高5%左右、耗时减少约80%。

中文关键词： 开源软件社区；软件网络；社会网分析；缺陷预测；软件度量

英文摘要： Collective intelligence (CI) emerges from the continuous on-line interaction between the masses and open-source software (OSS) communities, and promotes the rapid development of OSS. How to improve software quality with a scientific evaluation and prediction of CI's impacts on the development of OSS projects, is a challenging issue of current OSS basic research. The goals of our project are to (1) explore the inherent assocation between OSS's evolution (in terms of version update) and public praise (derived from social annotations) to reveal the mechanism of mutual influence and co-change between them; (2) propose a social annotation-driven crowdsourcing improvement method to increase the efficiency of triaging, reproducing and estimating a large number of software bugs; and (3) develop a prototype tool and conduct empirical studies to enhance the level of development and management of OSS projects. Our contributions are (1) the discovery of new features and mechanisms of OSS's evolution driven by public praise instead of traditional customer satisfaction, and a time series analysis-based prediction model for the potential of an OSS project; and (2) a bug tossing graph model based on Markov chain and collective expertises of groups within a larger developer community in crowdsourcing, whose accuracy is increased

英文关键词： open-source software community；software network；social network analysis；defect prediction；software measurement