Modelling Non-Linearity in Economic Classification with Neural Networks

Hennie Daniels, Bart Kamp, and William Verkooijen
Tilburg University
B.Kamp@kub.nl

Abstract

In this paper we present results of a study on economic classification with neural networks. A comparison is made between neural networks and linear modeling techniques, and we investigate the problem of overfitting and the estimation of prediction errors in cases where the available data sets are relatively small. it is shown that selecting network parameters by k-fold cross-validation combined with weight decay training is an effective remedy for those phenomena. The conclusions are illustrated in two cases: modeling the housing price in Boston and the classification of bond ratings.