Research On Commodity Futures Price Forecasting And Portfolio Optimization Based On Deep Learning

Although China’s commodity futures trading started late,it has a very large scale,and the volume of trading has been in the forefront of the world for nine consecutive years,so the investment strategy of commodity futures has attracted much attention.Firstly,based on the futures price index trading volume,wavelet decomposition subsequence and other characteristic sequences,this paper uses the deep learning model-LSTM to model and forecast the trend of five types of futures price index.In terms of prediction characteristics of time series,this paper decomposes the original sequence by Nonextractive Haar wavelet transform,so as to steadily decompose the long-term and short-term features contained in the price index series.Empirical research shows that the price forecasting model constructed in this paper performs better in the downward market environment,and performs worse in the upward market environment than in the downward market.Secondly,according to the forecast results,we construct the corresponding multi-warehouse and short-warehouse.Through the annual return rate,maximum withdrawal,Sharp ratio and other indicators,we find that for these five types of single futures,the strategy in this paper has better performance than the benchmark strategy-turtle trading strategy.However,in the aspect of risk control,the model in this paper is inferior to the turtle trading strategy.Finally,because the futures market has margin trading mechanism compared with the general securities market,the futures leverage is introduced to construct a Futures Portfolio Management model,which combines five types of futures based on the forecast results.The results show that,compared with single futures,the Futures Portfolio constructed in this paper has better performance.Moreover,the portfolio strategy of commodity futures effectively disperses risks,controls the maximum withdrawal of strategies,and significantly improves Sharp ratio,effectively remedies the deficiencies and shortcomings of this strategy in risk control.