Research On Optimization Of Adaptive Density Partition Clustering Algorithm

DPC algorithm is a clustering algorithm based on density division.Compared with traditional algorithms,it has the advantages of simplicity and efficiency,insensitive to noise,no iterative operation and clustering performance not affected by the dimension of data space.In recent years,it has attracted the attention of researchers.However,it also has some defects.Aiming at the problems that clustering by fast search and find of density peaks does not consider the internal structure difference of data when calculating the local density,the cluster center cannot be adaptive,the chain reaction when allocating sample points,and the high time complexity when solving the local density and relative density,many measures are proposed.The following is the main research content and innovation of this paper.Aiming at the problem that the existing DPC algorithm can not adapt the clustering center,a clustering algorithm with adaptive clustering center is proposed.By introducing relative density,and removing the outliers in the data set,selecting the core points and potential clustering centers to judge the connectivity between potential clustering centers.At the same time,aiming at the chain reaction caused by a point allocation error in the existing density peak clustering algorithm,which leads to the unsatisfactory clustering results of subsequent points,a density peak clustering algorithm based on minimum spanning tree is proposed.Firstly,the cluster center and core point are determined,and then the vertex of t he minimum spanning tree is made by the core point.Then calculating the Euclidean distance between all vertices,and the Euclidean distance is regarded as the weight of the edge of the minimum spanning tree.Taking any cluster center as the starting point,the minimum spanning tree is constructed by prim algorithm,and then the clustering is completed.This algorithm solves the problem of chain reaction caused by a point allocation error in the clustering process,and improves the accuracy of clustering results.Experimental results show that this algorithm can effectively solve the problems of adaptive clustering center and chain reaction of traditional clustering by fast search and find of density peaks.Aiming at the high time complexity of DPC algorithm in solving parameters such as truncation distance,local density and relative density,a density peak clustering algorithm based on grid screening is proposed.The concept of grid is introduced into the algorithm.Firstly,the data set is divided into multiple grid cells of equal width according to the specified parameters.Then the grid is divided into original group,secondary group and tertiary group to decrease the quantity of sample points involved in the calculation of local density and relative dens ity of sample points,so as to reduce the time complexity.The results of experiment show that compared with the traditional clustering by fast search and find of density peaks,DPC algorithm can effectively improve the calculation speed without reducing the calculation accuracy.