Link Prediction Based On Information Diffusion In Microblog Media

Hundreds of millions of users active in micro-blog media every day, theseusers can produce a large amount of information, many users have to face a mass ofinformation that they can t fast and efficiently find the information they interestedin. The users in micro-blog media follow others to receive information, so, they willfollow others if the information released by them is they interested in.The problem of link prediction in micro-blog media is fundamentally theproblem to provide users with a customized information service by recommendingfriends to them. In this article, we want to settle the problem of link prediction bythe analysis of the flow of information and the network structure. According to theflow of information, we put forward three link prediction algorithms, the first onebased on node similarity, the second one based on classifier model, the third onebased on random walk model.The link prediction algorithm based on node similarity by the informationdiffusion is the improvement solution of the recommendation algorithm based onnode similarity in traditional social networks, this algorithm calculate nodesimilarity by combine the flow of information, common followers and commonfriends, and choose the one who has the highest similarity with the target user torecommend to him. The link prediction algorithm based on classifier model by theflow of information hope to classify the users by the analysis of theinformation diffusion and the network structure, and choose the suggested userscategory as the recommended collection. The link prediction algorithm based onrandom walk model and information diffusion calculate the probability of the targetnode walk to other node, then sort the recommendation list according to theprobability value. From the experimental results can be seen that the supervisedlearning algorithm is the best, the next is the algorithm based on node similarity, thelast one is the algorithm based on random walk model. And the research workproved that the link prediction algorithm based on information diffusion is moreefficiency than the traditional methods, that proved the information diffusionfeature will be of great importance in link prediction field.