The SOM Methodology

Conventional statistical methods are able to reveal regularities, trends and structures in raw data. Few methods allow to directly visualize relations between elements in large and complex data sets. In this book we have provided several applications of a method introduced around 1982 called the Self-Organizing Map (SOM). The relations between data items become explicit in the SOM due to a nonlinear projection from a high-dimensional data space onto a two-dimensional display. As demonstrated, this method has been found useful for financial, economic and marketing applications. In this chapter we explain the SOM concepts and methodology, starting from concepts that are supposed to be generally known.