Ce. Imrie et al., River flow prediction using artificial neural networks: generalisation beyond the calibration range, J HYDROL, 233(1-4), 2000, pp. 138-153
Artificial neural networks (ANNs) provide a quick and flexible means of cre
ating models for river flow prediction, and have been shown to perform well
in comparison with conventional methods. However, if the models are traine
d using a dataset that contains a limited range of values, they may perform
poorly when encountering events containing previously unobserved values. T
his failure to generalise limits their use as a tool in applications where
the data available for calibration is unlikely to cover all possible scenar
ios.
This paper presents a method for improved generalisation during training by
adding a guidance system to the cascade-correlation learning architecture.
Two case studies from catchments in the UK are prepared so that the valida
tion data contains values that are greater or less than any included in the
calibration data. The ability of the developed algorithm to generalise on
new data is compared with that of the standard error backpropagation algori
thm. The ability of ANNs trained with different output activation functions
to extrapolate beyond the calibration data is assessed. (C) 2000 Elsevier
science B.V. All rights reserved.