PREDICTION INTERVALS FOR ARTIFICIAL NEURAL NETWORKS

The artificial neural network (ANN) is becoming a very popular model f or engineering and scientific applications. Inspired by brain architec ture, artificial neural networks represent a class of nonlinear models capable of learning from data. Neural networks have been applied in m any areas, including pattern matching, classification, prediction, and process control. This article focuses on the construction of predicti on intervals. Previous statistical theory for constructing confidence intervals for the parameters (or the weights in an ANN), is inappropri ate, because the parameters are unidentifiable. We show in this articl e that the problem disappears in prediction. We then construct asympto tically Valid prediction intervals and also show how to use the predic tion intervals to choose the number of nodes in the network. We then a pply the theory to an example for predicting the electrical load.