Clustering in the identification of space models

Abstract

Clustering is the process of grouping a set of objects into clusters so that objects within a cluster have high similarity with each other, but are as dissimilar as possible to objects in other clusters. Dissimilarities are measured based on the attribute values describing the objects (Han & Kamber, 2001). Clustering, as a data mining technique (Cios, Pedrycz, & Swiniarski, 1998; Groth, 2000), has been widely used to find groups of customers with similar behavior or groups of items that are bought together, allowing the identification of the clients’ profile (Berry & Linoff, 2000). This chapter presents another use of clustering, namely in the creation of Space Models. Space Models represent divisions of the geographic space in which the several geographic regions are grouped accordingly to their similarities with respect to a specific indicator (values of an attribute). Space Models represent natural divisions of the geographic space based on some geo-referenced data. This chapter addresses the development of a clustering algorithm for the creation of Space Models – STICH (Space Models Identification Through Hierarchical Clustering). The Space Models identified, integrating several clusters, point out particularities of the analyzed data, namely the exhibition of clusters with outliers, regions which behavior is strongly different from the other analyzed regions.