Analysis And Prediction Of Information Dissmination Based On User Behavior

With the development of Web 2.0,microblogs,blogs and other colorful social media continue to emerge and mature.The emergence of these social media provides the possibility of sharing resources,exchanging information and makes the internet to maintain substantial amounts of user behavioral data.These data are rich and valuable.How to mine useful information from massive data has become a popular research.We can acquire an understanding of the distribution of public behavior based on analysis of user behavior in the network,and this can assist in detecting abnormal behavior and provide reasonable bases for the study of public opinion control and influence evaluation.This thesis discusses the topic dissemination and the popularity for topic from two levels of user individual behavior and user group behavior by modeling,analysis and research.The contribution of this thesis can be summarized as follows:1.In the aspect of user individual behavior and the law of information dissemination,using user information,relationship and behavioral data from hotspots in social network,features that affect user behavior are extracted and the model of information dissemination prediction based on user retweeting behavior is counstructed to analysis the rules of information dissemination.Firstly,we propose a tensor-based mechanism for mining user interaction.At the same time,we can analyze the influence of the following relationship on the interaction between users based on the characteristics of the tensor in data space conversion and projection.Secondly,a time decay function is introduced for the tensor to quantify further the evolution of user behavior in current social hotspots.The function can be fit to the behavior of a user dynamically.Finally,we invoke time slices and discretization of the topic life cycle and construct an information dissemination prediction model based on logistic regression.In this way,we can explore the law of information dissemination in social hotspots.2.In the user group behavior and the situation of information dissemination,the model of topic popularity prediction is proposed using the participated users and their relationship.The model can predict the popularity of different stages of the topic.And it is helpful to control the evolution of the topic.Firstly,the data of social hotspots is serialized to form sequence.A grey prediction model is constructed using the advantage of grey system theory in mining valuable information based on partially known information.The model can predict the popularity for social hotspots.Also,it can achieve the correct description and effective monitoring of the development situation of social hotspots.At last,the experimental verification is carried out with the data set of Tencent micro-blog.Experimental results show that the model that we proposed can effectively combine the topology of information dissemination network and the characteristics of user behavior.It can not only intuitively analyze the characteristics that affect user behavior,but also explain the internal and external dynamics genesis of information dissemination.In addation,this thesis tries to explore the characteristics and laws of user behavior.