用户网络访问行为的预测研究

发布时间：2018-07-01 15:24

本文选题：社交网络 + 时间间隔　；参考：《中北大学》2017年硕士论文

【摘要】：随着Web2.0技术的推广应用,涌现出各类在线社交网站。借助这些社交网站,用户可以分享内容、表达观点、建立私密关系等,因而社交网站对丰富人们的情感、文化和娱乐等需求起到了很好的作用。与此同时,用户在社交网站上也留下大量的行为痕迹。基于这些行为信息,挖掘用户在线行为规律,预测用户在线行为,对舆情分析、网络安全、社会安全、信息推荐和商品营销等领域都具有极其重要的意义。本毕业论文就是要利用某高校用户对社交网络的访问数据,分析用户的社交网络访问行为特性并揭示用户在线行为背后的内在机理。具体研究工作及贡献包括:(1)基于个体层面和群体层面分析了用户连续两次在线访问行为之间的时间间隔分布,研究了用户时间间隔序列的相关性和活跃性等特征,并揭示了其背后的内在机制;(2)分析了用户访问行为中的记忆特性,发现用户的在线行为具有较强的短记忆性,其分布服从高斯分布。并据此建立了马尔科夫过程模型,用于解释用户访问行为中的记忆特性;(3)基于上述所发现的访问行为特性,本文进行了用户访问行为的时间序列预测研究。针对用户历史访问数据,采用ARIMA模型和Holt-Winters三参数指数平滑法对点击流时间序列进行预测分析,通过建立模型来预测数据的未来走向,并分析比较两模型的优劣。
[Abstract]:With the popularization and application of Web 2.0 technology, various online social networking sites have emerged. With the help of these social networking sites, users can share content, express their opinions, establish private relationships and so on, so social networking sites play a good role in enriching people's emotional, cultural and entertainment needs. At the same time, users on social networking sites also leave a lot of behavior traces. Based on these behavioral information, it is of great significance for the analysis of public opinion, network security, social security, information recommendation and commodity marketing to excavate the rules of online behavior of users and predict the online behavior of users. The purpose of this thesis is to analyze the characteristics of users' social network access behavior and reveal the underlying mechanism of users' online behavior by using the data of users' access to social networks in a certain university. The specific research works and contributions are as follows: (1) based on the individual level and the group level, this paper analyzes the time interval distribution between two continuous online access behaviors, and studies the characteristics of the correlation and activity of the user time interval series. It also reveals the internal mechanism behind it. (2) We analyze the memory characteristics of user access behavior and find that the online behavior of users has strong short memory and its distribution is distributed from Gao Si. The Markov process model is established to explain the memory characteristics of user's access behavior. (3) based on the characteristics of user's access behavior, the time series prediction of user's access behavior is studied in this paper. The Arima model and Holt-Winters' three-parameter exponential smoothing method are used to predict and analyze the click-stream time series. The future trend of the data is predicted by establishing the model, and the advantages and disadvantages of the two models are analyzed and compared.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【参考文献】