the neural networks appear to perform well. The training session for “Jasmine” rice was the most successful, with an MAPE of about 2%! The other two categories of rice –
“White” and Parboiled”– had average forecast errors of 3.2% and 3.8%, respectively.