Research On Social Relationships Based On Node Behavior

With the continuous development of WeChat,micro-blog and other social media,social networks have been integrated into people's life,are closely related to daily life,and gradually become an integral part of people's life.Social network is an important source of information exchange and acquisition,In the study of social networks,the most fundamental problem is the study of the relationship between users Therefore,it is of great practical significance to analyze user behavior and study the relationship between users,This thesis mainly studies the relationship between the users in the social network,by analyzing of the behavior of the nodes in the social network,the research of this thesis can be divided into the following two aspects:1.A key driving factor mining model is proposed.Considering the characteristics of the maximum entropy model,we do not need to take the independence of the features into account.First of all,based on the node interest,the relationship between friends,as well as the impact of community factors that affect relationship are generated.Then,aiming at the problem of the uncertainty of the weight distribution of each influence factor,the influence factor function is established from three aspects: the node interest,the friend relationship,the community influence and so on.Finally,based on the maximum entropy model,the key driving factors mining model is established,and the key driving factors affecting the user relationship are explored.The experimental result confirms that the model can not only mine the key factors that affect the user relationship,but also can effectively predict the user relationship2.A user relationship analysis model based on semi supervised learning is proposed.First of all,through the analysis of online social network data,the factors affecting the strength of the user relationship are extracted from the two aspects of the node's attributes and structural features.Then,the user relationship analysis model is established by using Tri-Training based on the semi supervised learning.Finally,the model is used to explore the strong links in social networks.Based on the semi supervised learning algorithm,we can use the unlabeled samples to train the classifier with good performance in the case of sparse samples.The experimental results confirms that the proposed model has higher performance compared with the traditional algorithm.In summary,the mining model and analysis model of user relationship study in this thesis,has greatly enriched the study of customer relationship and provided reference for further exploring of the development mechanism of social network structure.