Model selection in neural networks

In this article, we examine how model selection in neural networks can be g uided by statistical procedures such as hypothesis tests, information crite ria and cross validation. The application of these methods in neural networ k models is discussed, paying attention especially to the identification pr oblems encountered. We then propose five specification strategies based on different statistical procedures and compare them in a simulation study. As the results of the study are promising, it is suggested that a statistical analysis should become an integral part of neural network modeling. (C) 19 99 Elsevier Science Ltd. All rights reserved.