Research On The Price Discovery Function And Trend Forecast Of Shanghai Crude Oil Futures

Crude oil,as a non-renewable strategic resource,is the trading product with the strongest political attributes in the commodity futures market.The dominant position of crude oil pricing affects the country’s energy security and economic security.This article takes the closing price sequence of Shanghai crude oil futures as the main research object.First,we explore the price discovery function of Shanghai crude oil futures towards Daqing crude oil spot.According to the Granger causality test,there is a long-term equilibrium relationship between them;the Shanghai crude oil futures is the Granger cause of the spot price of Daqing crude oil,which is unidirectionally guided;the shock brought by the Shanghai crude oil futures market dominates the changes of Daqing crude oil price,and it has a certain degree of price discovery function.On the basis of passing the market effectiveness test,we construct different time series forecasting models to analyze the long-term trend of the Shanghai crude oil futures and guide current price trading.First,we built a classic ARIMA model as the benchmark.The ARIMA model has excellent in-sample fitting effects,and forecast errors within 1 to 3 trading days are also very small.It is a good choice for short-term investors,but not applicable to long-term trend prediction.The Prophet model is introduced into the prediction of futures price series,and the HWKS algorithm and sliding window are used to optimize the change point selection of the Prophet model.This paper proposes an improved HW-Prophet model,which can maximize its advantages in the selection and reconstruction of mutation points and reduce the forecast error by about 22%.At the same time,we use multiple market indicator to construct the LSTM model and optimize its parameters.The prediction performance of the LSTM model is the best in the single model,but it is very dependent on the adjustment of the hyperparameters according to the experimental design.HW-Prophet and LSTM have their own advantages in different prediction intervals.Finally,we make different combinations of LSTM and HW-Prophet models based on residual correction and optimal weighting.Since the residual sequence predicted by the HWProphet model is similar to white noise,it is difficult for LSTM to learn effective dynamic laws from it.The result shows that the combined method of residual correction is not suitable for the two single models in this paper.By minimizing the error sum of squares to solve the optimal weight coefficients and combining the two models linearly,the prediction advantages of the single model in different intervals can be effectively transmitted,and the prediction accuracy is greatly improved.The HW-Prophet-LSTM combination model constructed in this paper based on the optimal weight is currently one of the best models for predicting the long-term trend of Shanghai crude oil futures.