Post neural network procedure to determine expected prediction values and their confidence limits

The wider use of artificial neural network (ANN) analysis within some disciplines such as economics, could be considered to be impeded to some degree by the absence of confidence intervals for the resultant predictions from such networks. This paper outlines a post-neural network procedure that may be used to: (a) Determine whether or not the network is producing predictions that are statistically significant (b) Provide a mechanism to determine the confidence limits around the expected predictions of a neural network.