The effect of neural network parameters on the performance of neural network forecasting

This paper deal first with artificial neural networks for demand forecasting, neural networks have successfully been used for demand forecasting, however, due to a large number of parameters to be estimated empirically, it is not a simple task to select the appropriate neural network architecture for a demand forecasting problem. Researchers often overlook the effect of neural network parameters on the performance of neural network forecasting. This paper examines the effects of the number of input and hidden nodes and hidden layers as well as the size of the training sample on the in-sample and out-of-sample performance. The second objective of this paper is to describe a new forecasting approach inspired from regression method for weekly demand forecasting, we have used this approach for demand forecasting as a benchmark for comparison. This method performs an extensive search in order to select the appropriate transformation functions of input variables, the weighting factors and the training periods to be used, by taking into consideration the correlation analysis of the selected input variables. With this procedure the best forecasting model is formed.