The Research Of Community Detection Algorithm Based On Few Selected Nodes

In recent years, with the increasing amount of complex networks, and the expansion of fields involved in complex networks, the study of complex networks already become a well-known subject. Complex networks usually have a certain community structure:the internal relation closely and the external relation sparsely between communities, mapping to a complex network diagram, which internal edges are much and boundary less. How to quickly and effectively achieve community structures from a complex network is the direction of most of scholars and also has the important application value.Through access to a mass of the Chinese and foreign literatures, and the research of existing community detecting methods, two kinds of new community detection algorithms are designed: hierarchical agglomeration community detection algorithm via few selected nodes(HCDFSN) and active semi-supervised community detection algorithm via few selected nodes (ASCDFSN). HCDFSN algorithm introduces the measurement of node similarity and code nodes, and the method of processing overlap nodes. The main idea of HCDFSN is starting from the core nodes to build a new society until it is finished with all the core nodes, and then deal with overlapping between communities, and finally, the final community structure is obtained from the initial communities depended on modularity. Through the initial community forming method and community boundary processing method, the HCDFSN can solve the following two problems showing hierarchical agglomerative algorithm:(1) community boundary node once wrong, not repartitioning problem; (2) the effect of hierarchical aggregation algorithm is not ideal. The ASCDFSN suggests the acquisition method of score and the candidate nodes, and the active semi supervised strategy. The primary idea of ASCDFSN is to obtain the candidate set by the score acquisition method and then selects a part of nodes from them, artificial markers, and finally use the idea of semi supervised community detection. This algorithm can solve the following three problems by a small selected meaning nodes and the active semi supervised strategy faced with semi supervised algorithm:(1)randomly selecting a small number of nodes leads to poor effect of community detection problem; (2) the nodes chosen could not cover all community in semi supervised algorithm.This paper experiments on the multiple real network datasets and a manmade dataset, and the running process of algorithms could be visualized. Through experiment results show HCDFSN and ASCDFSN algorithm on community detection have good effect and high efficiency, compared with the same type methods.