A comparison of forecasting models using multiple regression and artificial neural networks for the supply and demand of Thai ethanol

This paper presented three types of models for forecasting the supply and demand of Thai ethanol, so called MR, ANN, and MR-ANN models. MR models were formulated using stepwise multiple regression analysis, which were statistically significant. However, MR models provided low performance in forecasting. ANN models were constructed using artificial neural networks, which provided satisfactory results. Moreover, the third type of models was an integration of multiple regression analysis and artificial neural networks. In MR-ANN models, influential factors from stepwise multiple regression, were taken as inputs for artificial neural networks. The integrated models provided a fair results comparing to the first two types of models. In summary, ANN models provided the lowest MAPE and the highest R² indicating that the models were the most appropriate among the three types of models. ANN models are therefore recommended to forecast the supply and demand of Thai ethanol.