基于LDA模型的微博人群粗粒度划分

摘要

Today's social network has a large number of users and the ensuing mass of information, which can provide important data for the study in various ifelds. Among them, how to divide the social network based in interest groups efifciently and accurately has become a hot research. This paper introduces the LDA model which commonly used in text analysis to solve this problem and provides some optimization.%当今的社交网络拥有庞大的用户数量和随之而来的海量信息，可以为工业界各个领域的研究提供重要的数据补充。其中，如何高效准确地对网络人群进行按照兴趣领域的划分成为了研究的热点。本文通过引入文本分析中常用的LDA模型，挖掘出微博用户和其单向关注的其他用户之间暗含的兴趣信息，建立主题模型来给用户进行粗粒度划分。同时本文研究了引入非对称超参数、弱连接理论、TF-IDF调频以优化本算法划分的正交度，并给出了相应分析。