Research On Quantitative Investment Strategy Of Stock Index Futures Based On VMD-ML Price Prediction Model

Price prediction is the basis of quantitative timing strategy.However,because the price of financial products is affected by many aspects such as macro fundamentals,policies,and group psychology.So improving the accuracy of financial time series forecasting has always been a difficult and hot research issue in the time series forecasting problem.Aiming at the two major difficulties of financial time series forecasting-instability and nonlinearity,this paper introduces variational modal decomposition and machine learning algorithms,respectively,and proposes a VMD-ML price prediction model.First,by introducing the variational modal decompositio,the problem of non-stationary financial time series is solved.Secondly,through the introduction of multiple machine learning algorithms for comparative analysis,to solve the limitation of a single machine learning algorithm in time series prediction.On the basis of the price prediction model,this article proposes a corresponding quantitative investment strategy and further optimizes the investment strategy from the perspective of optimization of the prediction model parameters and position management.First of all,this paper constructs the main continuous contract of the Shanghai and Shenzhen 300 stock index futures(referred to as IF9999),and conducts statistical and stationary analysis of the IF9999 price series.Secondly,this paper constructs a VMD-ML price prediction model,and conducts a prediction analysis based on the IF9999 logarithmic return sequence to verify the effectiveness of the prediction model.The price prediction model is mainly divided into three steps: The first step is to decompose the price time series through VMD to obtain multiple eigenmode components with different physical properties.The second step is to construct a machine learning(SVR,LSTM)prediction model for each component sequence.The third step is to obtain the final prediction sequence of the main continuous contract yield of the Shanghai and Shenzhen 300 stock index futures through reconstruction.Finally,this paper builds an investment strategy based on a price prediction model for backtesting,and optimizes the investment strategy from the perspective of hyperparameter optimization of the prediction model and position management.The experimental results of this paper show that:(1)Compared models,the VMD-ML price prediction model proposed in this with the SVR and LSTM paper can obtain smaller prediction errors(MSE,MAE and SMAPE).(2)Genetic algorithm can effectively optimize the penalty coefficient C in the SVR algorithm and the parameter gamma in the Gaussian kernel function.Grid search can effectively optimize the number of neurons and the number of iterations prediction model,a stock in the LSTM algorithm.(3)Based on the VMD-ML price index futures investment strategy that significantly exceeds the increase of the CSI300 can be constructed.