Research Of Community Detection Algorithm Based On Natural Nearest Neighbor

Many complex systems in the natural and social, can be described as a network ordiagram in a certain way. These networks have a complex internal structure, so calledcomplex networks. In recent years, with the complex network’s widely used in manyfields, the community detection and analysis in the complex network also get more andmore attention. The characteristic of community structure is that the nodes in thecommunity share a state of high density accumulation Associations, at the same time,the nodes between different communities share a state of low density accumulationAssociations, and the basic function of community structure detection is to take thenodes into the right community. In-depth study of the problem will help us study theentire network module, function and evolution in a divide and conquer approach.So far, many algorithms have been proposed to detect the community structure ofcomplex networks. The typical community detection algorithms can be divided intotraditional hierarchical clustering and graph segmentation algorithm, suchKemighan-Lin algorithm and spectral bisection algorithm, and GN algorithm andNewman fast algorithm. These algorithms have their own advantages and disadvantages,has its specific scope, and also needs further improvement. Some algorithms can beimprove in computational efficiency, and some algorithms can be improve in algorithmthinking. However, the overwhelming majority of these algorithms divide the networkinto several mutually separated communities and accuracy depends on the selection ofinitial conditions, and can’t recognize the important nodes in the network. In2011, XuXiaowei proposed community structure detection algorithm based on a similar structure.The advantage of this algorithm is that it not only divide the community structure of thenetwork, and can identify a particular node in the network. But the algorithm is verysensitive to the choice of parameters, it is difficult to ensure the quality of the detectedcommunities.In this paper, inspired by the SCAN algorithm, borrowed the idea of structuralsimilarity, used the natural nearest neighbor algorithm without parameters, we propose anew community detection algorithm-community detection algorithm based on naturalnearest neighbors. Meanwhile, neighbors concept itself contains a community, whichmakes natural nearest neighbor algorithm can be a good way to sort out the nodes in thenetwork. In this paper, we improved the concept of original natural nearest neighbor by using structural similarity as the basis for determining the neighbors. Then we give adetailed analysis of the advantages and disadvantages of various natural nearestneighbor graph and propose a suitable construction method to detect communities withnatural nearest neighbor graph. Finally, we experimented with real data sets, andcompared with the source data and results of the SCAN algorithm, proved the feasibilityand effectiveness of the algorithm.