一种基于共享近邻的密度峰值聚类算法

基于加权K近邻的密度峰值发现算法(FKNN-DPC)是一种简单、高效的聚类算法,能够自动发现簇中心,并采用加权K近邻的思想快速、准确地完成对非簇中心样本的分配,在各种规模、任意维度、任意形状的数据集上都能得到高质量的聚类结果,但其样本分配策略中的权重仅考虑了样本间的欧氏距离。文中提出了一种基于共享近邻的相似度度量方式,并以此相似度改进样本分配策略,使得样本的分配更符合真实的簇归属情况,从而提高聚类质量。在UCI真实数据集上进行实验,并将所提算法与K-means,DBSCAN,AP,DPC,FKNN-DPC等算法进行对比,验证了其有效性。

Clustering Algorithm Based on Shared Nearest Neighbors and Density Peaks

Robust clustering by detecting density peaks and assigning points based on fuzzy weighted K-nearest neighbors(FKNN-DPC) is a simple and efficient clustering algorithm,which can automatically detect the cluster center and assign the non-cluster center sample based on weighted K-nearest neighbors quickly and accurately.It is powerful in recognizing high quality cluster in any scale ,any dimension,any size and any shape of the data set,but the weight calculation in assigning strategies only considers the Euclidean distance between samples.In this paper,a similarity measure based on shared neighborhood was proposed,and the sample assigning strategy was improved by this similarity,so that the cluster is more consistent with the real attribution,thus improving the clustering quality.The effectiveness of the algorithm is verified by comparing the experiments on the UCI real data set with the K-means,DBSCAN,AP,DPC,and FKNN-DPC algorithm.