Análisis socio-económico del Area Metropolitana de Valencia por medio de un Mapa Auto-organizado de Kohonen

Resumo

This paper is about the possibility of applying Kohonen's neural model to the study of a Region¿s internal structure. After showing the main characteristics of the working and learning of Kohohen models with uni-dimensional (L.V.Q.) and bi-dimensional (S.O.M.) exit, it is stated that the first -applied to the analysis of a territory- can be an interesting alternative to the statistical methods of cluster analysis. On the other hand, the Self Organized Maps (S.O.M.) can be of richer grading, as they result in a clustering of the territorial entities about categories placed in a map (bi-dimensional typology) - which gives the possibility of incorporating topological criteria into obtained clustering. After these considerations an empirical application is carried out: starting from the socioeconomic information available about Valencia ¿s Metropolitan Area towns. We go on to group them using these models and some traditional methods of Cluster study. The quality of the different groupings is compared according to the internal homogeneity of those obtained. And, lastly, the results are interpreted by different statistical and topological analysis (Manova/Discriminant).