Neural Networks in Regional Researches

Abstract

Electronically storaged datas and documents boost in an accelerating way in nowadays Information Society due to the revolution in computer technology. Nevertheless there is no time and manpower to process these datas by conventional manual analysis. This problem has been solved with some new analytical methods applied in the past few years. In general they are called data mining methods. In this article I give a brief summary on one of these methods: the artificial neural networks. The first part of this essay gives a review about artificial neural networks and their usage in GIS and geography. In the second part I describe the clusterization of Hungarian small regions with a Kohonen Self-Organized Map (SOM). The SOM has made four clusters in the analysis. After the investigation, I labelled them as capital city, urban, urban-rural transition and rural regions. The clusters were homogeneous and can be compared against previous reserches such as ESPON 7 „Urban-rural relations" project. In comparison there was a high level similarity between the outcome of the above clusterizations. Finally, I point out that Kohonen SOM ís an acceptable method of grouping in geographical researches.