Density Clustering with a Combined Distance as a Tool for the Analysis of Urban Environment

Abstract

A new method for clustering spatial data, based on distance, generalizing the difference density of the distribution of the quantitative trait and spatial distance is offered. Two methods to select the best number of clusters are considered, one of which is based on the nonparametric density estimation and correlation ratio. As an illustration a comparison of the results of cluster polarization of data for Saratov in 2006-2007; algorithm is realized as a set of functions for the statistical package R, available on request to the authors.