Forecast And Analysis Of Futures Copper Price Based On Wavelet Transform

With the industrial production of non-ferrous metals copper has a more extensive need, and increasing production and consumption of copper, copper futures also becoming an increasingly important means of hedging and investment.But the actual market hides a variety of confounding factors, frequent fluctuations in copper futures market, resulting in price movements are difficult to predict, making the risk of future dealing also growing bigger and bigger. Therefore, decision of production and operation is right or wrong, depending on the transition of the market’s demand can not be properly grasp, in particular, If properly grasp market movements in the next step, and scientific and rational way to predict copper futures prices, so you can help companies to take advantage of supply and demand situation of the market, to avoid the risk of price fluctuations from copper futures. to improve enterprise production and management decision-making’s rationality and scientific,truly get to quota according to need, thereby enhancing economic efficiency of enterprises. Advise on decision-making valuably to investors in futures and spot demanders.This paper is based on a large number of mainly related literature review and ananlysis of the situation on the futures market in recent years to carry on writing and analysis. I select that variables mainly from three aspects of having significant influence on copper futures prices: the copper fundamentals, industry level, macroeconomic to predict and analyze copper futures’ s tendency of price, to establish the model of copper futures’ s price prediction based on wavelet transform.Bockus-Jenkins once said: "For non-stationary time series built regression model directly is spurious regression, the results of which will be reduced to get the credibility." Well first of all, for raw sequence data and low-frequency via on wavelet transform’s decomposition and synthesis do the ADF unit root test and cointegration test. Because non-stationary time series under the same order integration condition exist a long-term stable relationship, according to the result of unit root test and cointegration test, we get the same order integration as Shanghai copper futures price index variable. Then refer to the important variable choosing of random forest to get the same order integration and cointegrated that important for dependent variable, then according to the result of vector auto regression model to determine the optimal lag order,and finally through each AIC, SIC, R ^ 2 of the establishment of a number of different regression models’ s lag order and the significance of the estimated coefficients of the variables to determine which is the best prediction model of the original sequence copper futures prices and after wavelet transform low frequency sequence. Establish ARMA model for high-frequency sequences of synthesized by the wavelet transform, and predictive the value of the wavelet transform,then synthesis the predicted of low and high frequency to do a short-term predict. Thus we compare the predictive value of a single model and combined for short-term, the results show that the copper futures prices combined forecasting model based on wavelet transform compared to a single copper futures price prediction model can make better short-term price of copper futures index forecast.