Soybean Commodity Futures Timing Strategy Based On Wavelet Analysis And AR-BP Network Prediction Model

Since the reform and opening up,China's futures market has developed well,and the trading volume and types of futures products are also increasing rapidly.As futures have certain function of venture investment,a large number of investors are pouring into futures market continuously.At present,in the field of commodity futures research,an important research direction is how to build a quantitative investment strategy based on futures price prediction model,so as to help investors avoid risks and improve investment returns.This paper selects the representative soybean futures as the research object of quantitative timing strategy.After a comprehensive study of the research status of quantitative investment strategy at home and abroad and the principles and characteristics of futures price prediction methods,this paper proposes the construction of a time selection strategy based on the wavelet analysis of AR-BP neural network prediction model for futures price prediction.After obtaining the time series data of soybean commodity price,the wavelet analysis was used to de-noising and smoothing the data.The coif4 wavelet basis and the number of three-layer decomposition layers were selected to determine the threshold rules and threshold adjustment methods of Rigrsure and sln,respectively,and then used in the case to obtain the low and high frequency signals of soybean price data.ARIMA time series model and BP artificial neural network model were built for two kinds of low and high frequency signals respectively for data processing and prediction.Finally,the results of low and high frequency prediction are recombined and applied to the timing strategy for verification.By comparing different models,this paper finds that the original time series model is not accurate enough to predict the futures price,resulting in a low winning rate in the actual trading.After continuous improvement and optimization of the model,a time selection strategy based on the wavelet analysis AR-BP neural network model was constructed.The trading opportunity was similar to that before the improvement,but the winning rate was significantly improved,and the indicators such as annualized return,sharpe ratio and maximum withdrawal were improved.This paper discusses the feasibility and effectiveness of the application of the time selection strategy based on the wavelet analysis AR-BP model in futures investment,and provides some reference value for investors in investment choice.However,due to the diversification of futures product types and the complexity of influencing factors,there is still a long way to go to build a timing strategy that can accurately predict futures prices.