Using Spatial Autocorrelation Measures for Data Clustering
Eduardo Morales, Yosu Yurramendi Mendizabal · Dialnet (Universidad de la Rioja) · 2007
Unsupervised classification or clustering has been used in many disciplines and contexts. Traditional methodologies are mostly based on the minimization of the distance between data and the cluster means without considering any other possible relationship present in data, e.g. spatial interactions. We propose a clustering evaluation function based on a measure of spatial autocorrelation and show its application to clustering. We provide some examples of the quality of the proposed criterion.