Research on the Optimized Clustering Method Based on CFSFDP

Aiming at deficiencies of the clustering by fast search and find of density peaks algorithm (CFSFDP) in which cluster centers cannot be selected automatically and the segmentation of cluster points is unreasonable, this paper proposes an optimized CFSFDP algorithm (O-CFSFDP) based on kernel density estimation and the max-min clustering algorithm. First, kernel density estimation is used to adaptively determine the optimal value of the cutoff distance by iteration. Then the max-min clustering algorithm is adopted to obtain the number of categories of a given data set, and by calculating the product of local density and distance for each point, corresponding cluster centers can be selected. Finally, O-CFSFDP introduces the idea of local density within the cluster. By computing it, points at the edge of each cluster will be aggregated more reasonable. Contrast experiments show that O-CFSFDP can not only realize cluster centers automatic selection and reasonable partition of cluster points, but also achieve better performances in clustering robustness and accuracy.