Modeling gasoline demand in land transport sector in Iran by using GMDH neural network

Gasoline is one of the main energy carriers in land transport sector. With regard to the low price for gasoline in Iran, the study of the non-price factors which affect gasoline demand is important.
In this paper, for identifying the most important factors which affect on gasoline demand, GMDH neural network model as a tool with high capacity in the diagnosis of non-linier complicated processes especially with limited number of observations is used.
At first, in a primary model, the effects of 10 variables which are titled inside and outside of system variables, on the demand for gasoline during 1387-1355 are evaluated. Then by using deductive process in the three stages, 5 variables identify as more important variables.
In the final stage, the variables such as gross domestic product per capita, the numbers of vehicles (with gasoline fuel), the gasoline subsidies and the foreign exchange rate in unofficial market have double and liquidity has normal influence on gasoline demand.
In addition to the
In addition, the results show that the importation of variables such as foreign exchange rate, liquidity and gasoline subsidies in the model increase the model efficiency.