Density Peak Clustering Algorithm Based on High Density Connection with Entropy Optimization

Abstract

In this paper, an MM-HDC (Max Mean and High Density Connection) method was proposed to find the initial clustering center based on the maximum mean distance and fuse each cluster based on the high density Connection. Firstly, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\Delta\rho=70\%$</tex> was set to select the initial clustering centers and the mean distance was introduced. The selection of cluster centers will be stopped until the distance between the desired new mean center and some previously selected cluster centers is less than <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$2^{\ast}d_{c}$</tex> , and the selection of initial cluster centers is completed. Then use the allocation policy of k-means to clustering all the data points by the distance between each initial clustering center and data points, constantly updated after cluster center, center for migration, until the old and the new cluster centers position changed little (the distance is very small), then stop update clustering center, and the last of the clustering results as the final clustering results. Finally, iterative fusion method is used for center fusion to get better clustering results. Experimental results of classical data sets show that the MM-HDC method is superior to the DPC algorithm and k-means algorithm, and the improved density peak clustering algorithm has higher accuracy. Moreover, The MM-HDC algorithm can obtain satisfactory results on the data set with special shape or uneven distribution.