Density peak clustering improved by quantity particle swarm optimization

The Density Peak Clustering (DPC) algorithm demonstrates unique advantages in handling complex data. However, its traditional distance measurement and manual selection of the cutoff distance are limited, leading to biased local density estimation and unstable clustering results. To address these issues, this paper proposes an improved DPC algorithm based on Quantitative Particle Swarm Optimization (QPSO-DPC). Firstly, the Mahalanobis distance, which considers the covariance structure of data, is introduced to accurately characterize the similarity between data points. Secondly, the global optimization ability of PSO is utilized to automatically search for the optimal cutoff distance, avoiding subjective bias. Experiments show that PSO-DPC significantly improves clustering accuracy on multiple datasets, demonstrating superior performance.