Density Peak Clustering Analysis Based on Coulomb Force Similarity Metric

Density peak clustering, as an important type of clustering analysis method, has several advantages, including the ability to quickly identify cluster centers, the lack of need for prior knowledge or iterative processes, and the use of few parameters. However, existing density peak clustering methods use a single-step allocation strategy, which results in poor clustering performance, particularly for high-dimensional data leading to the "curse of dimensionality". The Coulomb force and the propagation of cluster center labels are studied, and a density peak clustering analysis method is proposed. First, Coulomb force is employed to redefine local density and relative distance, effectively alleviating the issue of distance concentration in high-dimensional data. Then, a k-d tree is used to search for the neighbors of each point, which effectively realizes the propagation of cluster center labels, improving clustering performance. Based on this, a density peak clustering analysis based on coulomb force similarity metric algorithm is proposed. Finally, experiments on high-dimensional UCI datasets validate that the proposed algorithm achieves good clustering performance and effectively mitigates the issue of distance concentration, making it suitable for high-dimensional clustering tasks.