Density Estimation Clustering Method Based on Reverse Nearest Neighbor
The density peak clustering algorithm based on mutual nearest neighbors(DenMune)is an effective clustering method that calculates the local density of data points through mutual nearest neighbors.However,this algorithm has the problem of unreasonable construction of clustering skeletons,and using a hard voting strategy when allocating weaknesses can easily lead to errors.Therefore,a new density estimation clustering algorithm based on reverse nearest neighbor(RNN-DEC)is proposed.This algorithm introduces reverse nearest neighbors to calculate the local density of data points,dividing them into strong points,weak points,and noisy points.It builds the skeleton of the clustering algorithm using strong points and assigns weaknesses to the cluster with the highest similarity through soft voting.Finally,a cluster fusion algorithm based on reverse nearest neighbor is proposed,which fuses high similarity sub clusters to obtain the final clustering result.The experimental results show that compared to other classical algorithms,this algorithm has better clustering performance on some synthetic datasets and UCI real datasets.
万方数据
维普
