This paper analyzes recursive and rolling neural network models to forecast one-step-ahead sign variations in gold price. Different combinations of techniques and sample sizes are studied for feed forward and ward neural networks. The results shows the rolling ward networks exceed the recursive...