The Problem Of Modularity And Its Algorithm Research In Community Detection

With the rapid development of information technology, people gradually find that there exits common characteristic in the emerging networks-such as protein network, power network, social network, neural network-which we called complex network. These networks are in close relationship with human daily life, the network studies can not only rise living standards but also promote the development of human beings, therefore, it is of great significant to research complex networks.Community structure is an important way to understanding Community Algorithms. It appears many kinds of Community Detection Algorithms as the development of research on community structure of complex networks. Module Maximization Algorithms is an important part of these community detection algorithms, when compared with other algorithms, it shows that this algorithm can better evaluate the result of algorithms because of the ability of qualifying Community Detection Algorithms’ result. However, its modularity has the problem of resolution limit which lead algorithm to make small communities combine together during calculation and get incorrect result. In response to these issues, this article proposes a new algorithm to calculate density of nodes in network and transferred result into module local optimum Community Detection Algorithms which improves accuracy of algorithm.k-path centrality is introduced in this article to calculate the density of nodes and prepossess network, it can both effectively balance centrality of every side in the network and reflect the network global information. First, K-value will be optimized to reduce calculating time and the density between nodes will be also calculated by using centrality of every side and node degree. The density between nodes combines the global network information and the local network information will to better reflect relationships between the nodes. Then, community detect the combination of the matrix of density and LM Algorithm to get the result. At last, implement this algorithm, an experiment was made between artificial network and real network to find out the calculating time and results of different algorithms in different K-value and offer reference when selecting k-value in network and at the same time compares with other module maximum algorithms. The experiment result proves module degree of algorithms has increased and be able to discover more communities, which effectively solved the resolution limit problem.