Advanced clustering analysis for environmental indicators

Abstract

Maps are used in many application areas to support the visualization and analysis of geo-referenced data. The geometry used in those maps is usually associated with the administrative subdivisions of the regions, disregarding the purpose of the analysis. Another common drawback is that traditional classification methods for data analysis, used for example in Geographic Information Systems, divide the data in a pre-defined number of classes. This can lead to a situation where a class integrates values that are very different from each other and that do not allow the identification of the main differences that can exist between regions. This paper presents a different approach for geo-referenced data analysis that is based on clustering analysis. Through a clustering process it is possible to analyse a specific data set with a map, employing a Space Model, which suits the purposes of such an analysis. Space Models are new geometries of the space that are created to emphasize particularities of the analysed data.