跨社会媒体的用户行为分析与协同建模方法

One social media is insufficient to model users’ multifaceted and complex behaviors in social media context. Thus user behavior analysis and modeling across different social medias has begun to be emphasized in Computer Supported Cooperative Work (CSCW) and Social Computing studies and some other related areas. For the problems and challenges of current research on user behavior analysis and modeling across social medias, this project focuses on the key scientific problem-computing integration of user behavior characteristics’ cooperation between social medias, and studies user behavior analysis and cooperative modeling methods. First, we will propose a social understanding and computation expressing integrated method for user behavior analysis and understanding across social medias, which combines the theories and methods of social science as well as methods and techniques of computer science to analyze and understand users’ expectations, motivations, intentions as well as decision-making and implementation process of content posting and social relationship building by utilizing multiple social medias, and then investigate their behavior characteristics related to content posting and social relationship building across different sites. Based on that, we will propose a cooperative modeling method for user generated content across social medias and a cooperative modeling method for social relationship strength across social medias. In the end, Weibo and Douban will be used as research sites to evaluate the above methods by utilizing content recommendation and friend recommendation as applications.

基于单个社会媒体难以实现对用户在社会媒体环境下行为特征的多侧面和系统刻画，跨社会媒体的用户行为分析与建模研究开始受到计算机支持的协同工作与社会计算等领域的关注。针对当前跨社会媒体的用户行为分析与建模研究存在的问题及面临的挑战，本项目围绕“社会媒体之间用户行为特征协同的计算融合”这一关键科学问题，研究跨社会媒体的用户行为分析与协同建模方法。首先，提出社会理解与计算表达交叉融合的跨社会媒体的用户行为分析与理解方法，融合社会科学的理论与方法和计算科学的方法与技术分析理解用户使用多个社会媒体进行内容发布和社会关系构建的期望、动机、目标及行为决策与实现过程，揭示用户在多个社会媒体环境下内容发布和社会关系构建的行为特征；在此基础上，提出跨社会媒体的用户生成内容协同建模方法和跨社会媒体的社会关系强度协同建模方法。最后，以微博和豆瓣为研究站点，以内容推荐和好友推荐为应用，对上述方法进行验证评估。

本项目针对当前跨社会媒体的用户行为分析与建模研究存在的问题及面临的挑战，研究了跨社会媒体的用户行为分析与协同建模方法，主要创新成果包括：提出了社会理解与计算表达交叉融合的跨社会媒体的用户行为分析与理解方法；在分析理解用户使用不同社会媒体进行内容发布和社会关系构建的行为决策与实现过程和行为特征的基础上，提出了跨社会媒体的用户生成内容协同建模方法和跨社会媒体的社会关系强度协同建模方法；并以微博和豆瓣等为研究站点，以个性化推荐为应用，对上述方法进行了验证评估。在本项目资助下，共发表学术论文13篇，主要创新成果发表在高水平国际会议和期刊上，其中CCF A类会议/期刊论文5篇，CCF B类会议/期刊论文4篇，中科院SCI一区期刊论文1篇；同时申请国家发明专利1项；培养4名硕士研究生毕业，协助培养1名博士研究生毕业；组织国内学术研讨会3次，参加国际学术会议8人次。