A Crude Oil Price Forecast Method Based On Subsystem Modeling And Forecast Combination

The fact that crude oil holds an important and indispensable position in the global economy determines the necessity for crude oil price forecast. The world economy is now in a critical period of recovery and China's international status is also enhanced greatly, both of which determine the urgency of crude oil price forecast. Price of WTI is forecasted in this paper through the full application of three levels of Forecast Combination: combining forecast information, combining forecast methods and combining the forecast results. At first, five main factors influencing crude oil price are extracted, which are named supply factor, demand factor, stock factor, economy factor and finance factor. Each factor is considered as a subsystem. Then five forecast sub models are built in five subsystems respectively. In the subsystem modeling process, firstly the variables of top importance are filtered out by Granger Causality Test, and then two classical forecast methods (Causality Model and Autoregresstion Model) are combined to forecast the price of crude oil. On evaluating the forecast accuracy of five subsystems, hypothesis testing is originally occupied, which outperforms the traditional evaluation methods in the empirical analysis of this paper. Finally, the forecast results of these five subsystems are combined to act as the ultimate forecast result, which is more robust than any of those five ones. Apart from the quantitative analysis, a "Crude Oil Prosperity Index" is fabricated according to the performance of some certain variables in this paper, which is the combination of forecasting information. It is verified by empirical research that the Crude Oil Prosperity Index is doing well in predicting the price trend. The results of quantitative analysis and qualitative analysis are integrated to present a relatively full view of the international oil market.