The Density Peaks Clustering Algorithm Based on a Combination of Polynomial Kernel and EP Kernel with K-Nearest Neighbors

To address the problem of inaccurate cluster center selection in the Density Peaks Clustering (DPC) algorithm when applied to datasets with non-uniform density distributions, this paper proposes an enhanced clustering algorithm called PKEPK-DPC. The proposed method introduces a polynomial kernel-weighted distance metric, and integrates both the Exponential Power (EP) kernel function and the K-Nearest Neighbors (KNN) algorithm to construct a novel local density estimation approach. This combination effectively improves the algorithm's ability to capture similarities among high-dimensional and nonlinear data points. Experimental results on benchmark datasets demonstrate that the proposed algorithm achieves superior performance in terms of clustering accuracy and cluster center identification compared to existing methods.