Influence Model Building And Link Prediction Algorithm In Social Networks

With the rapid development of information technology and the popularization of intelligent devices,social network has become the most extensive virtual platform that affects public life.Great changes have taken place in people' s traditional pattern of working,studying and living.However,there are huge challenges in mining high-value information from actual social network and applying it rationally,due to the social network having diversity and complex relationships.In the related work about social network,there are two hot issues.On the one hand,it adopts the key users mined by influence model to realize the public opinion guidance,on the other hand,it uses the link prediction algorithm to solve information overload problem.In view of the two hot issues,the main research contents can be concluded as follows:1.Influence model of social networks based on popularity index.When analyzing and studying social network,it is of great significance to distinguish the high-value potential users by influence model both in academic research and business application.Taking Sina Weibo for example,this paper propose a WLRank algorithm,which adds the popularity index after eliminating the cheat users in the network.This algorithm assigns the weight value of the nodes according to the popularity index,which result in a faster convergence of the algorithm.We compare WLRank algorithm with classic PageRank algorithm and LeaderRank algorithm over Sina Weibo,Sina Forum and DBLP datasets.In addition,we use four common central measurements to evaluate the results,which verifies that WLRank algorithm performs better than existing classical methods both in accuracy and rationality when mining potential high-value nodes in the network.2.Social network link prediction algorithm based on meta-path weights.As the core of the recommendation system,the link prediction algorithm can solve the problem of information overload in the network effectively and improve the user experience.Aiming at the problem of most link prediction algorithms in simple homogeneous social network,we propose a META-Base link prediction algorithm based on meta-path weight.This algorithm establish a binary heterogeneous network model in a real social network and use the mete-path to describe the association between nodes in the network.Our algorithm uses the instance path to calculate the weight of the meta-path,and adopts the the probability of the link to predict the link relationship between the nodes.We conduct extensive experiments on Sina Weibo,Sina Forum and DBLP datasets.We use the link prediction algorithm based on common friends,common topic and similarity of topic as the comparison algorithms.The final result shows that the accuracy rate and recall rate of the META-Base algorithm are better than other algorithms,and the result of the experiment is more reasonable.