An Improved K-means Clustering Algorithm

Hui Xu, Shunyu Yao, Qianyun Li, Zhiwei Ye

202030 citationsDOI

Abstract

Among the existing clustering algorithms, K-means algorithm has become one of the most widely used technologies, mainly because of its simplicity and effectiveness. However, the selection of the initial clustering centers and the sensitivity to noise will reduce the clustering effect. To solve these problems, this paper proposes an improved K-means clustering algorithm. The concept of CLIQUE grid is used to remove the noise and obtain the regional density. Then the initial center point is selected according to the method of Fast Search and Find of Density Peaks (CFSFDP). And the influence of grid density error on the selection of initial center point is reduced by the idea of granularity, while the selection of cluster center by manual participation in the peak density algorithm is then avoided. Compared with the original K-means algorithm, the improved algorithm proposed in this paper has higher precision, smaller difference in clustering effect for different data, and lower parameter dependence.

Topics & Concepts

Cluster analysisCURE data clustering algorithmCorrelation clusteringComputer scienceSelection (genetic algorithm)Data stream clusteringCanopy clustering algorithmAlgorithmGranularityNoise (video)Data miningDetermining the number of clusters in a data setk-medians clusteringFuzzy clusteringGridPoint (geometry)MathematicsArtificial intelligenceGeometryImage (mathematics)Operating systemAdvanced Clustering Algorithms ResearchData Mining Algorithms and ApplicationsData Management and Algorithms