Adaptive Density Peak Clustering with Optimized Border-Peeling

Density Peak Clustering (DPC) is good at processing datasets with irregular shapes and uneven density. However, the traditional DPC algorithm has limitations in identifying cluster centers and accurately describing the distribution of data points, including parameter dependence and insufficient processing of border points. To solve these problems, this paper proposes an adaptive density peak clustering algorithm. This algorithm uses an optimized border-peeling strategy to divide the dataset into core points and boundary points, thereby improving cluster center identification and data distribution characterization. In addition, a novel density kernel calculation method based on the concept of natural neighborhoods is introduced, which provides an efficient estimate of local density for datasets with uneven density distribution and eliminates potential errors caused by manual parameter selection. Finally, a two-step allocation strategy based on fuzzy logic principles is designed to enhance the allocation process of the DPC algorithm, thereby improving the accuracy and efficiency of clustering. Experimental evaluations on synthetic and real datasets demonstrate the superior performance of our algorithm compared to existing methods.