Research On Clustering And Community Detection Algorithm Based On The Neighborhood Information

With the wide application of data mining technology in many areas,researchers pay more and more attention to the related algorithms in this field.Clustering and community detection,which are essentially the similar,are two important topics in the field of data mining.Clustering refers to dividing the data set into a number of subsets(which are called clusters),so that the data objects within the same cluster are similar,while the data objects from different clusters are dissimilar.In addition,community detection can be regarded as the extension of clustering idea in network data.It refers to dividing the nodes in the network into several communities,so that the nodes within the same community are similar or closely linked,while the nodes from different communities are dissimilar or with sparse connections.So far,the related research works have done a lot of work on clustering algorithm and community detection.In this paper,after deeply exploring the role of neighborhood information in clustering,a new clustering algorithm NIDD is proposed for the shortcomings in the existing clustering algorithm(i.e.FDP).Then,the core idea of NIDD is extended to community detection in the social network,and the fuzzy centrality and fuzzy membership based on neighborhood information are introduced.Then,a novel community detection algorithm FCFM is proposed.Finally,the fuzzy relation based on neighborhood information is studied,and another community detection algorithm CDFR based on the idea of FDP is proposed.This paper mainly includes the following three aspects:(1)A clustering algorithm NIDD based on neighborhood intersection and density difference is proposed.After analyzing the classic clustering algorithm FDP,we find that it is not effective in some data sets.In view of the shortcomings of FDP,we propose a novel clustering algorithm NIDD.The idea of NIDD is:First,in the process of clustering expansion,if the reference point and the point to be expanded belong to the same cluster,then the density difference between them should be relatively small,otherwise,if they come from different cluster,the density difference should be relatively large.Second,if the intersection of the k-nearest neighbors of the reference point and the point to be expanded is small,then they should be divided into different clusters.Finally,we demonstrate the effectiveness of the clustering algorithm NIDD with experiments.(2)A non-overlapping community detection algorithm FCFM based on fuzzy centrality and fuzzy membership is proposed.We will discuss in depth the role of fuzzy centrality and fuzzy membership in community detection.The main idea of FCFM are as follows:First,we introduce the fuzzy centrality and fuzzy membership based on neighborhood information.Second,the node with the largest fuzzy centrality in a community is considered to be the central node of that community.And the community begins to expand from that central node by the use of fuzzy membership.We compared the experimental results of FCFM and several classical community detection algorithms on several real network datasets,and the results verify the effectiveness of FCFM.(3)A community detection algorithm CDFR based on fuzzy relation is proposed.First,we proposed the concept of NGC node.And then,the algorithm of calculating the fuzzy relation between each node and its NGC node is given.The fuzzy relation we proposed in this paper can be regarded as the degree of dependence on its NGC node.A smaller degree of dependence indicates that the node has large autonomy.In other words,the node is more likely to be a central node of the community.Finally,which community a node belongs to depends on its NGC node.It is worth noting that,although our algorithm is based on fuzzy relation,it is different from the fuzzy overlapping community detection model.Most of the community detection algorithms involving fuzzy concepts are used to discover overlapping communities.However,CDFR is a non-overlapping community detection method based on fuzzy relation.This paper not only has a reference value in the research of clustering algorithms,but also has reference value of community detection in social network.