Manifold Density Peaks Clustering Algorithm

Abstract

Cluster analysis is a popular technique in statistics and computer science with the objective of grouping similar observations in relatively distinct groups generally known as clusters. In this paper we propose an approach called Manifold Density Peaks Clustering to improve the basic density peaks clustering. It mainly concerns three aspects. First, geodesic distance is adopted to calculate manifold distance matrix. Then, cluster centers are identified automatically by the number of clusters, which improves the inconvenient way of identifying cluster centers by dragging a rectangle manually. Moreover, isometric mapping is introduced to map high dimensional datasets into lower dimension, so as to achieve the goal of dimensionality reduction. Experimental results indicate that the proposed algorithm has achieved the expected results on both low and high dimensional datasets.