A Density-Peak-Based Clustering Method for Multiple Densities Dataset

Clustering methods in data mining are widely used to detect hotspots many domains. They play an increasingly important role the era of big data. As advanced algorithm, density peak clustering (DPC) algorithm is able deal with arbitrary datasets, although it does not perform well when dataset includes multiple densities. The parameter selection cut-off distance dc normally determined by users’ experience and could affect result. In this study, a density-peak-based method proposed clusters from datasets densities shapes. Two improvements made regarding limitations existing methods. First, DPC finds difficult Each cluster has unique shape interior different This adopts step merging approach solve problem. Second, high points can automatically be selected without manual participation, which more efficient than methods, require user-specified parameters. According experimental results, applied various performs better traditional DPC.