Algorithm Of Detecting Community Structure On Weighted Networks

Community structure is an important property of many real complex networks. Search for and analysis of community structure is conducive to understanding of network structure and master network information better. Since community structure has been discovered the algorithm has become a hot topic research in the field of complex networks. So far as it development, there has been a variety of fast and accurate algorithms to detect community structure. But it is still have contradictions between time complexity and accuracy. That is to say,the algorithm with lower time complexity will get unsatisfactory result, while the accuracy algorithm has higher time complexity. So it's difficult for us to analysis large-scale network reliability of community structure. Most algorithms are detecting community structure for the network with Boolean relations.In fact, in most reality network, there are some objective information can not be ignored between nodes. Most real networks are weighted network. Therefore, it is very necessary to design an algorithm that could solve the contradictions between time complexity and accuracy, and could analysis community structure on weighted network.Based on the above issues, this article introduces the algorithms of community structure, weighted network with community structure modeling and improve the CNM algorithm test on computer-generated network, then take the stock market as an example to analysis community structure of weighted networks. The results show that the issues raised in this paper can be well resolved, and get good results.The main contribution of paper is as follows:1. Take a bottom-up method to solve the loss of small-scale community structure and the network nodes can't be correct classification during the process of detecting large-scale network community structures. That is to improve the accuracy of the algorithm.2. At the algorithm, consider the alternative of edge could reduce the storage space and improve the speed of algorithm operating.So it avoid "fence node" error classification.3. Improved CNM algorithm, introduce nod weight and link weight to make the algorithm applicable to detect community structure of large-scale weighted network, then use it to analysis stock market price volatility.