Evaluating the density parameter in density peak based clustering

Abstract

The density peak based clustering algorithm is a simple yet effective clustering approach. This algorithm firstly calculates the local density of each data and the distance to the nearest neighbor with higher density. Based on the assumption that cluster centers are density peaks and they are relatively far from each other, this algorithm isolates the candidates of cluster centers from the non-center data. After the cluster centers are identified, the other data are assigned labels equaling to those of their nearest neighbors with higher density. In this way the clustering can be accomplished efficiently and clusters of arbitrary shapes can be obtained. The key of the density peak based clustering algorithm lies in the density calculation method. In this paper we study the influence of the data amount used in density calculation on the clustering results of the density peak based algorithm. As a result, we arrive at some conclusions on the selection of the data amount, which can be useful in applying this algorithm in real tasks.