Improve Self-organizing Combination Forecasting Model Emd In The Oil Futures Market

Oil is one of the most important energy resources. The fluctuation of the oil price directly affects the development of a country's economy, while the oil futures prices are the expectation of spot oil prices. Therefore, the accurate forecasting of the oil futures prices should be the critical area concerned by government, enterprise and researchers. However, the oil futures prices involve the typical characteristics of time series data, nonlinearity and nonstationarity, which brings insuperable difficulties in the price forecasts. Researchers in an increasing number find that certain potential information or pattern exist behind the plausibly ruleless price fluctuation in the financial markets. Especially, in the oil futures market affected by many factors, extracting information imbedded in the data to study further will provide effective guidance to investors.Specific to the characteristics of oil futures prices data, this paper conbined the EMD (empirical mode decomposition) approach with the existing forecast methods, and proposed a Self-Organising AC (Analog Complexing) forecasting model improved by EMD, which then is used in the empirical study. First we employ EMD method to decompose the crude oil prices in NYMEX, and obtain a series of intrinsic mode function (IMF), which are stationary and of strong periodicity, and a trend term. Then we determine the mode length based on the periodicity of each IMF, and make the 1-step and 3-steps out-of-sample rolling forecast using AC arithmetic. The empirical results indicate that predictive ability of this model is upstanding.In order to improve the accuracy of forecasting model, this paper draws the self-organizing data mining (GMDH) into the forecasting model, to build GMDH combined forecasting model improved by EMD. We use GMDH method to construct the optimized intelligence combinations for the forecast of IMF and trend term. The result shows that combined forecasting model based on GMDH extracted the characteristics of data abundantly and displayed its unique superiority.In the basis of data characteristics in the oil futures prices, this paper draws EMD methods into the build of forecasting model, extracts the imbedded information in the data, combines existing forecasting methods with data mining approach to construct the forecasting model, demonstrates that this model is of upstanding predictive ability through empirical study and provides feasible advices for the investors of oil futures. Moreover, this model also has favorable generalization ability and thus can be applied in the forecast of other products in financial markets.