WCSE

A Clustering Algorithm Based on Find Density Peaks

Abstract— A clustering algorithm named “Clustering by fast search and find of density peaks” is for finding the centers of clusters quickly. The algorithm has the advantages of fast clustering speed and simple realization. But after several experiments on the algorithm,we found that the algorithm results excessively depends on the selected cluster center, if a cluster has multiple density peaks or multiple clusters share the same density peak leads to incorrect clustering, and the algorithm will be determined too many points as noise. In view of the disadvantages, a new way proposed to optimization of CFSFDP by using clusters local density distribution graph and hierarchical clustering algorithm,and identify noise according the outlier degree of the point (DMCFSFDP). Firstly, the new algorithm used CFSFDP algorithm to determined cluster centers and clustering data set. Secondly, DMCFSFDP division clusters based on local density distribution graph and merged the clusters that could be merged by using improved hierarchical clustering algorithm. Finally, the algorithm uses the outlier degree to identify the noise. The results of experiments have shown the DMCFSFDP algorithm is more effective than CFSFDP algorithm in clustering.

Index Terms— clustering, density peaks, local density distribution, merge clusters, otulier degree

Wang Peng, Wang Junyi
1School of Software, Inner Mongolia University, CHINA

[Download]

Cite: Wang Peng, Wang Junyi, "A Clustering Algorithm Based on Find Density Peaks," Proceedings of 2017 the 7th International Workshop on Computer Science and Engineering, pp. 81-85, Beijing, 25-27 June, 2017.