Files
Abstract
Using the Artificial Neural Network model which is known as an excellent tool for perception of data patterns and short-run forecasting, this study develops an early warning index for the world grain markets. To obtain the best result, this study compares the indices which are computed from the different specifications of output neurons in our neural networks. The final result of this comparison shows that the best early warning index can be derived when we use the most aggregate variables and include all leading variables between present and targeted forecasting time as output neurons in our networks. We also verified that the index form of this study shows much better performances in the view of prediction power when we compare the index with the existing indices from other studies.