Research On Community Discovery Algorithm Based On Node Similarity

In recent years,with the rapid development of information technology,social networks,such as micro-blog,forums and other social media rapid development.People share knowledge,learning technology in the network,mutual communication between each other,the social network has become an important part of people's life.In the network,people are often viewed as network nodes,these nodes have universal connection,nodes with some common characteristics constitute a large or small online community,the research on them can be a good guide to real life,has the practical significance,the community detection has been a hot the study of social networks.The social network has a large number of members and complex connections.The traditional algorithm of relational circle mining has a high complexity,and the performance of the network is reduced on a large scale network structure.Compared with traditional community detection algorithm,tag propagation(LPA)has great advantage in time complexity,and its improved algorithm SLPA also has the ability of mining overlapping communities,but the inherent randomization strategy of label propagation algorithm makes the algorithm unstable.Based on the characteristics of complex networks,this paper makes use of node similarity and label propagation to carry out community discovery research.First of all,this paper studies the theory of node similarity,and chooses the most feasible method of node similarity measure as the research method in this paper.Secondly,according to the characteristics of SLPA algorithm,change the label's initialization process using the node similarity:(1)according to the degree of node arrangement them,and assigned a label to the node with maximum degree;(2)Calculating the similarity between the current node and the neighbor node,if the similarity of a node is greater than the average similarity of all the neighbors,distribution the same label with the current node with.After all neighbors are completed,repeat the above operations are according to the node order until all the nodes have labels.Finally,choose the label with the maximum probability to propagate in the propagation process;if there is more than one maximum weight label,calculate the similarity of nodes with maximum weight label to the node to be updated,select the label has the largest average similarity.Through the above steps,the randomness and instability of the original algorithm are reduced.The experimental results show that the improved algorithm is more stable and effective than SLPA.