Application of Kohonen networks for clustering of the Zachodniopomorskie Voivodeship districts in terms of the level of socio-economic development

Abstract

The author presents the possibilities of using artificial neural networks in a multidimensional analysis – cluster analysis. The empirical example using districts of the Zachodniopomorskie (West Pomeranian) Voivodeship is the illustration of theoretical considerations. The study used statistical data from many areas related to socio-economic development: demography, labour market, natural environment, recreation, culture, social and technical infrastructure, and the economy. The aim of the study was to divide the voivodeship into disjointed typological groups of districts using Kohonen networks (Self-Organizing Maps). Several networks differing in structure of the output layer were constructed and trained. Selected diagnostic features of socio-economic development of districts were their input values. Using verified Kohonen networks, various sets of groups of the researched objects were created, and confirmed them are a useful tool for identifying clusters of districts similar to each other in terms of the level of socio-economic development.