A Method Of Link Prediction In Multi-dimensional Social Networks

With the rapid development of the networks, the networks are becoming increasingly largeand complex. Looking for unknown link in relationships of the existing network to get whatpeople want to learn and cognitive. Many scholars pay attention to this area. The link predictionin network is a method to solve this problem.As far as link prediction in dynamic and multi-dimensional social networks is concerned,many questions are still in the exploratory stage. Link prediction in multi-dimensional socialnetworks aims at estimating the likelihood of the existence of links between nodes inmulti-dimensional social networks.In the process of designing link prediction algorithm, the choice of the similarity is a veryimportant issue in the link prediction system. In examining the relevant link predictiontechnology, based on similarities between nodes, this paper presents a link prediction algorithmin multi-dimensional social networks. The factors of weight and node are considered for thelink prediction, while the related technologies in multi-dimensional networks are used in theprocess of link prediction. Finally, the method of link prediction for dynamic multi-dimensionalsocial network is proposed in this paper. Superiority of the algorithm is analyzed by the theoryand experiment.Main works in the paper are as follows:Firstly, research on the characteristics of dynamic multi-dimensional social networks. Bydeep studying of complex network theory, combined with the definition of dimension of themulti-dimensional nature, characteristics of small-world network and characteristics of changesof nodes in dynamic characteristics, main features of dynamic multi-dimensional socialnetworks are analyzed and obtained.Secondly, build a model of dynamic multi-dimensional social networks based oncomplexity theory. After the understanding of the essential characteristics of the dynamicmulti-dimensional social network, through the integrated use of multi-dimensional networks,weighted networks, and dynamic networks modeling, dynamic multi-dimensional socialnetworks are modeled.Thirdly, propose a link prediction algorithm for dynamic multi-dimensional socialnetworks. Based on structural similarity and the algorithm of common neighbors, a linkprediction algorithm for dynamic multi-dimensional social network is proposed. The algorithmconsiders weight and network structure into link prediction algorithm and divides known edgeE into the training set of ETand the test set EP. The level of the pros and cons of the algorithm isdetermined by precision indicators. Finally, simulation based on the algorithm is designed. We have taken a comparativeexperiment. By the appropriate algorithm, a dynamic multi-dimensional social networks isconstructed, and finally through the experimental，the superiority of the link prediction methodsis proved.The experimental results prove the process of link prediction is not isolated, it should fullyconsider the inherent characteristics of networks. In this paper, based on the relevantcharacteristics of the dynamic multi-dimensional social networks, a targeted link predictionmethod for dynamic multi-dimensional social networks is proposed. The prediction methods inthis paper can forecast existence of links between nodes more accurately. Similar evaluationcriteria prove that the method is effective, and it can solve the link prediction problem in thedynamic multi-dimensional social networks.