Table 6 shows the validation errors of the artificial neural networks as percentages of the validation errors of the BoxJenkins
method. The ANN produced better predictive accuracies, except for “white rice” where the MAPE is 20% above the
Box-Jenkins method. For “Jasmine rice,” the MAPE of the ANN is less than half of the Box-Jenkins method. Even with the
hard-to-predict “glutinous rice” prices, the MAPE of the ANN is 75% of the Box-Jenkins method.