Performance analysis and forecasting on Crude Oil: Novel support vector regression application to market demand

Forecasting demand plays a major role in manufacturing and a great big influence in modern countries. A good forecast can save the cost, lack of demand and expenses of surplus. Existing of oil in the production chain in various manufactures shows its importance. Main Object of this paper is to forecast Iran oil consumption forecasting, by using Support Vector Regression algorithm, which is measured to determine the amount of its extraction to be sell based on market´s demands. Predicting the amount of oil extraction is correlated with some other important deliberations in that country. For such diversity, the utilized algorithm employs a neural network to estimate a function between social and economic inputs and oil consumption as output. GDP, population, import and export are the inputs and the vital considered parameters of this study. The required data regarding to the oil production and extraction is taken from OPEC bulletin. To accurately reach to our main aim, we used the data from 1987 to 2011, which the available data from 1987 to 2006 is used to train the algorithm and data from 2007 to 2011 for tests the capability. The results show the algorithm has 0.0005 errors for the last five years and the SVR algorithm can be more accurate than other development approaches in oil consumption forecasting and shows the aim significance.