The application of an Extended Self-Organizing Map Networks to Market Segmentation

Abstract

This paper presents the technique of an extended SOM networks and shows how it can be applied as a data analysis tool to market segmentation. The Self-Organizing Map (SOM) network, an unsupervised learning neural network, is a categorization network developed by Kohonen. In this research, we implemented an extended version of SOM networks that further groups SOM output map into user specified number of clusters. We then compared the extended SOM network and K-means analysis, a popular clustering technique, in the context of market segmentation problems. The a priori motivation for considering a neural networks alternative is that SOM networks, like most of the neural network models, do not assume the multivariate normality of data and hence, may be more robust. Test results indicate that the extended SOM networks perform better than K-means analysis when the data are skewed.