基于OTUCM模型的网络社团在线识别  

On-line Identification for Network Community Based on OTUCM

作　　者：翟菊叶[1,2] 张浩 

机构地区：[1]蚌埠医学院卫生管理系,安徽蚌埠233030 [2]复旦大学计算机科学与技术学院,上海200433

出　　处：《情报理论与实践》2018年第7期129-135,共7页Information Studies:Theory & Application

基　　金：安徽省高校人文社会科学重点项目"基于对比挖掘的医疗卫生网络舆情的发现;跟踪及倾向性分析"(项目编号:sk2015A405);安徽省高校人文社会科学重点项目"基于统计分析及多维关联规则挖掘的安徽省大学生手机依赖与身心健康关联性研究"(项目编号:sk2017A182);2018年度安徽高校人文社会科学研究项目"医院负面口碑补救行为对患者满意度影响与多维度关联分析"(项目编号:SK2018A1072)的成果

摘　　要：[目的/意义]通过挖掘网络社团及其演化过程,有助于掌握网络用户结构及行为方式,对网络舆情分析具有十分重要的意义,当前的网络社团演化挖掘模型存在着局部最优及抗噪能力差等问题。[方法/过程]通过建立网络社团与主题的对应关系,把社团的识别转化成主题的识别。在OLDA模型的基础上引入用户参数,提出一种主题社团混合模型(OTUCM),该模型把t-1时刻社团后验作为t时刻社团先验,通过建模用户、消息、主题、词汇四者之间的概率依赖关系,生成t时刻消息—主题、主题—词汇、主题—用户三个分布矩阵。通过交叉熵计算t与t-1时刻的主题相似度,得出主题及主题下用户的演化过程。[结果/结论]在实验数据集上验证了OTUCM模型,并与当前流行算法进行对比,实验结果表明OTUCM模型在网络社团演化挖掘方面达到了良好的效果。[Purpose/significance] Mining the network community and its evolutionary process helps to master the structure and behavior of network users,which is of great significance for the analysis of network public opinion. The current evolutionary mining model of network community has some problems,such as local optimal and poor anti-noise ability. [Method/process] The identification of community is transformed into the recognition of topic by establishing the relationship between the network community and topic. The on-line topic and user combine model（ OTUCM） is proposed by introducing user parameters into OLDA model. This model can generate the distribution matrices of messages-topics,topics-words and topics-users of the time t by taking t-1 time topic posterior as t time topic priori and modeling the depending relationship among the possibilities of users,messages,topics and words. Through calculating the topics similarity between t and t-1 by the cross entropy,the evolutionary process of topic and its users are obtained. [Result/conclusion] The OTUCM model is verified by using the experimental data set. Compared with the current popular algorithm,the experimental results show that the OTUCM model achieves better results in mining the network community and its evolutionary process.

关 键 词：网络用户 网络社团 OLDA模型 网络舆情 主题社团混合模型 演化过程 

分 类 号：G206[文化科学—传播学] G353.1