Density Clustering Analysis Algorithm Based On Variable Neighbor And Adaptive Density

Clustering analysis is one of the important research contents in the field of data mining.Its main task is to divide the data set into multiple cluster using the similarity between data objects,so that the similarity of data objects in the cluster is maximized,and differences between data objects among clusters are great.Density clustering is a kind of important clustering analysis method,which can realize fast clustering analysis to arbitrary shape,and don’t need any prior knowledge.However,the existing density clustering analysis requires artificially set parameters and cannot adapt to the transformation of data objects density in the clusters.In this thesis,the density clustering analysis algorithm based on neighbor points and adaptive density is studied deeply by making full use of the distribution features of the data objects neighbor points,which effectively overcome the shortcomings that density clustering algorithm parameters cannot adapt to the large differences in the datasets,improve the effect of parameters in different datasets,and effectively promote the effectiveness of the cluster analysis.The main research works are as follows:(1)A density clustering algorithm is proposed by using effective neighbor points and adaptive density distribution.Firstly,the telescopic radius is determined by the relative distance of the data object nearest neighbor points,and determined effective neighbor points of the data object,aiming to avoid selecting neighbor points influence of clustering effect.Secondly,core points and boundary points threshold are calculated by using the relative distance,so that core area objects in the cluster are determined according to the effective neighbor points,and uneven density distribution and large distance among clustering clusters are improved by adjusting the effective distance within the clusters.In the end,experimental results verify the effectiveness of the algorithm on artificial datasets and UCI datasets.(2)A density peak group clustering algorithm is proposed by using variable neighbor points.Firstly,data objects neighbor points information are used to determine variable neighbor points with high density similarity,aiming to reduce the influence of neighbor points on the judgment of clustering centers.Secondly,the candidate clusters that may have density peaks in the datasets are determined based on the reverse neighbor points and variable neighbor points of the data objects,and the class clusters of the dataset are judged and selected clustering centers by constructing a decision diagram.Finally,the remaining data objects are assigned based on the assignment rules of the density peak clustering(DPC)algorithm.The effectiveness of the algorithm is verified with artificial datasets and UCI datasets.