Research On Community Detection Method Of Complex Network

With the further development of intelligence-led police work,the role of data in public security work is becoming more and more important,and the data mining and analysis have become an important investigative measure.According to the actual needs of intelligence analysis of Public Security Organs,this paper discovers the potential complex network models of actual works by analyzing a large number of crime and Internet data.Through the research and analysis of complex network,the group relationship between personnel and network elements can be excavated,and the data analysis ability and case handling efficiency of public security organs can be greatly improved.In this paper,some new community detection methods are proposed based on the actual network model.The accuracy and validity of the methods are verified by comparing the results of the model and the results of actual cases.All the research results in this paper have been applied to the practical application system of public security,which obviously improves the intelligence analysis and mining ability of the public security business system.The main work and contributions of this paper are as follows:(1)A MBS community detecting method based on node similarity of multi-dimensional network is proposed.According to the characteristics of multi-dimensional network structure,a similarity model of multi-dimensional network nodes is proposed.In Community Detection Method Based on Maximum Modularity,we verified a suppose that “if the Q is the largest when the number of community is m then the Q becoming larger as the number of community k closed to m”,which based a modified binary search algorithm(MBS)is proposed.A new community detection algorithm based on the similarity of multi-dimensional network nodes,is proposed as the MBS algorithm is applied to community detection.The experimental results show that SMBS algorithm can calculate the optimal number of community partition in the shortest time,and partition the network into communities,and get the global or approximate global optimal partition results,with high accuracy and efficiency.(2)A bipartite network community detecting method base the local sensitivity is proposed.In this section,the local sensitivity is applied to the detection of community in bipartite networks,and some models of bipartite network,such as the related community,related barrel and related degree,are proposed.Based on the above concepts,the idea of local sensitivity and stack are introduced into the community detection of bipartite networks,and a community detection method of bcipartite networks based on local sensitivity is proposed.Firstly,according to the degree of the first type of node,the bcipartite network is divided into several communities.Then,according to the degree and modularity of the second type of node corresponding to the first type of node community,it is merged into the first type of node community corresponding to the maximum degree and modularity.The experimental results show that this method can greatly improve the efficiency of community detection,and the quality of community partition is higher than other conventional dichotomous network community detection algorithms.(3)A parallel and efficient clustering algorithm based on MapReduce is proposed.This algorithm obtains initial centroid sample set by collecting the centroid points which meet the conditions of triangle.The optimal clustering results are obtained by calculating the DB value on the corresponding sample set with different K values without setting the initial K value,also the optimal cluster number k and the corresponding initial centroid.Because of the optimal kvalue and initial centroid,the cost of time of this algorithm is much lower than that of the ordinary k-means algorithm.At the same time,the difficulty of processing large-scale data with improved k-means algorithm is solved,also the efficiency of the algorithm is improved,by using MapReduce technology to compute the improved k-means parallelly.The experimental results show that this algorithm can greatly improve the efficiency of processing large-scale data,and the quality of community partition is higher than that of ordinary K-means algorithm.