Modelling Relationship Influence Of Nodes In Heterogeneous Social Network

The influence between nodes in a social relationship network refers to the ability of a node's opinions,attitudes,and behaviors to influence or change other nodes' opinions,attitudes,and behaviors.Studying the quantitative relationship of mutual influence between nodes has a very important role in discovering key nodes in the network,community division,social relationship prediction,and the drivers of public opinion evolution.Due to the variety and huge amount of social network data,the mutual influence of the behavior between the nodes depends on many factors.How to discover the internal factors of the influence between nodes and calculate their weights,and finally accurately and reasonably quantify the construction of the influence micro model is a research difficulty at present.On the other hand,the current research has not analyzed whether perspective of relationship characteristics plays an important role in different social fields,but instead uses the results of behavioral interactions between nodes to reversely calculate the influence,which only reflect the short-term impact of the node in influence model.Therefore,it is necessary to build a micro-influence model to better predict the long-term impact on node behavior that analyzed from different types of interaction relationships.In view of the above problems,this paper proposes an inter-node influence analysis model.By analyzing many different factors between nodes,we measure and construct a quantitative analysis model of influence between nodes in heterogeneous social relationship networks to predict the degree of mutual influence of node behaviors and verify the effectiveness of the model in large networks.The main work and contributions of this paper are as follows:(1)In view of the current influence analysis based on historical data learning,lack of modeling the internal mechanism of action,it is difficult to portray the long-term effectiveness of influence.In this paper,node attribute information and different types of interactive information are used as heterogeneous nodes.The influence of the nodes in the social relationship network is studied by two types of information,and the heterogeneous nodes are constructed into factor graphs.The mechanism of factor action is analyzed through the belief propagation algorithm to establish the above-mentioned node information weight calculation model.An influence modeling method called AIIM based on node attributes and interaction is proposed,which can achieve more fine-grained quantitative analysis of the mutual influence between nodes.(2)In the calculation of node importance evaluation of social relationship network,in order to solve the problem that the importance of nodes cannot be accurately measured from the analysis of network structure or single attribute,a node based on influence analysis is proposed from the perspective of mutual influence between nodes using AIIM method,and then a node importance evaluation algorithm based on influence analysis called IEIARank is proposed.Using the constructed heterogeneous social relationship network dataset,a comparative experiment is carried out on the proposed IEIARank algorithm and two existing importance evaluation algorithms.The experimental results show that the ranking results of the node importance status given by the proposed algorithm are better,and at the same time,the algorithm has strong mobility.(3)For accurately predicting the behavior changing of nodes which are influenced by other nodes in heterogeneous social relationship networks,a research on the change of node behavior or viewpoints by information dissemination is carried out.Based on the AIIM method and the IEIARank algorithm from the perspective of influence between nodes and the durability of information dissemination,an influence dissemination model called ID-NPM based on different influences between nodes is proposed.Using the heterogeneous social relationship network data set constructed in this paper to verify the change of node research field,compared with the classical machine learning algorithm,the experimental results show that the proposed ID-NPM model is more accurate for predicting node behavior changes.At the same time,the model has a certain generalization ability and can be applied to social relationship networks in various types or fields.