基于位置的社交网络用户行为分析与研究

发布时间：2019-04-07 19:28

【摘要】：随着信息时代的发展,基于位置的服务已成为很多行业研究的目标,而社交网络已广泛成为用户分享与交流的平台。基于位置的服务与社交网络的结合为用户的行为分析提供了重要的数据来源,有利于提高用户的个性化服务质量。研究用户之间的相似性对分析个体用户历史行为以及预测其未来的行为有着重要的意义。但是由于数据的多样性,使得用户相似性分析以及行为预测成为业界的热点和难点。本文考虑到用户数据的多样性特点,以Foursquare数据为研究对象,分析并研究了基于位置的社交网络用户行为。首先,本文针对于结构化数据提出了一种基于矩阵分解的用户潜在特征挖掘算法,针对于非结构化数据采用了一种基于主题模型的分析方法,融合这两种信息,共同分析用户之间的相似性：然后,本文分析了语言模型与基于位置的社交网络之间的共同特性,提出基于语言模型的行为预测算法,结合用户之间的相似性以及其它特征,使用决策树来预测用户的下一个访问位置；最后,本文进行了实验分析和验证,实验证明了基于两种信息融合的相似性分析有着良好的效果,并对基于语言模型的行为预测有一定的促进作用。
[Abstract]:With the development of information age, location-based service has become the research goal of many industries, and social network has become the platform of sharing and communication widely. The combination of location-based services and social networks provides an important data source for user behavior analysis and helps to improve the personalized service quality of users. It is of great significance to study the similarity among users to analyze the historical behavior of individual users and to predict their future behavior. However, due to the diversity of data, user similarity analysis and behavior prediction have become hot and difficult in the industry. Considering the diversity of user data, this paper takes Foursquare data as the research object, and analyzes and studies the user behavior of social network based on location. First of all, this paper proposes an algorithm based on matrix decomposition for mining the potential features of users for structured data. For unstructured data, a thematic model-based analysis method is used to fuse these two kinds of information. Then, this paper analyzes the common characteristics between language model and location-based social network, and proposes a behavior prediction algorithm based on language model, which combines the similarity between users and other characteristics. The decision tree is used to predict the next access location of the user. Finally, this paper has carried on the experiment analysis and the verification, the experiment has proved that the similarity analysis based on the two kinds of information fusion has the good effect, and has the certain promotion function to the behavior prediction based on the language model.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09;TP391.1

【相似文献】