Using Social and Economic Indicators for Modeling, Sensitivity Analysis and Forecasting the Gasoline Demand in the Transportation Sector: An ANN Approach in case study for Tehran metropolis

Compared to the conventional methods, Artificial Neural Networks (ANN) are considered to be one of the reliable tools for modeling of complex phenomena such as demand. Aim of this study is to provide a model for gasoline demand in transportation section of Tehran metropolis through multilayered perceptron neural network and using the presented model in analyzing the sensitivity of the model to the input variables and forecasting gasoline demand. Seven social and economic indicators are considered on a monthly basis within 2010-2016: fuel price, population, median household income, Gini coefficient, hybrid/gasoline cars ratio, price index of goods and services, and average vehicles lifetime. The average percent error 3.8% and 4.6% for the training and test data were obtained respectively. Sensitivity analysis results showed that, hybrid/gasoline cars ratio SX6F= -2.580, population of Tehran SX1F= 1.596 and average lifetime of the vehicles SX7F = 0.698 have greater influence on gasoline demand in transportation section. Fuel consumption was predicted by three different scenarios of moderate, pessimistic and optimistic by 2022. The prediction results showed that in case of the continuation of the current trend of the descriptive variables of the model, gasoline demand in Tehran transportation section will be increased by 2022. Finally, the result of projection of the gasoline demand is presented based on three scenarios for future 48 months (4 years). These scenarios are based on the 3 baseline, pessimistic and optimistic views illustrating the demand of gasoline by 2020.