المستخلص: |
This paper deals with the application of support vector machine in financial time series forecasting. The paper investigates the causal relationship between disaggregated petroleum consumption and economic growth in Egypt using support vector regression. Petroleum consumption will be disaggregated into six main fuel types: Gasoline, Fuel Oil, Gas Oil, Kerosene, LPG (Liquefied Petroleum Gas) and Natural Gas. The results indicate that SVM provides a promising technique in time series forecasting. For natural gas model MSE in training set =0.005 and in validation set MSE=0.001. For gasoline model MSE in training set =0.003 and for validation set MSE=0.008. For gas oil model MSE in training set =0.004 and in validation set MSE=0.001. For fuel oil model MSE in training set =0.003 and in validation set MSE=0.002. For Kerosene model MSE in training set =0.003and in validation set MSE=0.01. Finally, for LPG model MSE in training set =0.006 and for validation set MSE=0.001, which indicate very high accuracy of the predicted models.
|