Determining the number of dimensions underlying customer-choices with a competitive neural network

More and more data about consumer-behaviour is becoming available through the use of database systems and electronic-banking systems. In this paper we describe a neural network (NN)-based technique to extract some useful information from a database holding data about customers of an insurance-company. Using this NN, we try to learn more about the needs and wishes that lead consumers towards specific behaviour, so that better service can be provided in the future. We formulate a more formal version of this problem, describe the NN used, and show results on artificial and real data