Signed Network Community Detection Algorithm

Many complex real-world systems can be used to describe the network, such as neural networks, social networks and so on are all complex networks. In recent years, It’s raise a boom both here and abroad about community detection. A social network refers to the interaction among members to form a relatively stable system of relations. Generally, these networks have a common characteristic of community structure. Community structure to divide the network as the size of a number of communities. The community structure of signed network to keep the community is mainly connected to the internal community is mainly between the negative connection. In this paper, we focus on the community detection algorithm of signed network and weighted signed network in social networks, and put forward a community detection algorithm which is suitable for signed network and weighted signed network.Firstly, analyzed the basic concept of the signed network, study the algorithms of community detection. And analyzed several classical community detection algorithms.Secondly, according to Tushar Sharma et al proposed weighted signed network community detection algorithm in the second phase of node clustering existed problems. Improved clustering algorithm does not consider the overall situation is not the neighbor node of the cluster nodes reunion type of problems, the improved algorithm improves the modules of the network, using modularity to evaluate the results of divisions.Again, the algorithm is applied in the signed network when choosing different nodes as the initial node, network ultimately divided the community into different structure. Namely, the algorithm is not stability, presented the initial node selection algorithm. When the algorithm is extended to the general positive whole network, the number of nodes may generate a lot of relatively small communities, so introduction of the cluster density and cluster connecting factor increased community divided, efficiency, results of divisions in the signed network more reasonable.Finally, through the simulation experiment of the improved algorithm is tested on real data set, accuracy the algorithm is verified and validity by the modularity, error rate and frustration value.