Grid Strategy Of Futures Intertemporal Arbitrage Based On Machine Learning To Predict Volatility

In China’s financial derivatives market,futures trading is an important component of financial derivatives trading.With the development of computer technology,quantitative trading has gradually become an important tool of futures trading.Investors profit from the futures market by constructing quantitative trading strategies.However,the prediction and judgment of financial indicators is the key to profit or not.Therefore,accurate prediction of financial indicators such as volatility is of great significance to financial institutions.This thesis creatively constructs a quantitative strategy combining the traditional quantitative strategy and statistical model,predicts the realized volatility through the neural network model,so as to judge whether the transaction is or not,and uses the traditional grid trading strategy to trade in the period with small market volatility,so as to make a profit.The LSTM model in machine learning is used to predict the realized volatility of futures spread.At the same time,the attention theory is innovatively introduced.The attention LSTM model is constructed by using the attention layer before the LSTM layer.Compared with the ordinary LSTM model and the traditional Volatility Prediction Model GARCH,it is concluded that the attention LSTM model is better than the ordinary LSTM model and the traditional GARCH model in the Volatility Prediction of futures spread.In the empirical analysis of the prediction of price spread volatility of rebar futures,this thesis selects the data every ten minutes to predict the data of the next minute.In order to prevent over fitting,dropout is added,and the generalization ability of the model is improved by adjusting parameters.From the results,the average percentage error and mean square error of attention LSTM model in the verification set are lower than those of ordinary LSTM model and GARCH model,indicating that the generalization ability of attention LSTM model is better in the prediction of Futures Spread volatility.In the subsequent quantitative strategy backtesting,the grid strategy constructed by attention LSTM has higher yield and smaller maximum pullback compared with the other two models.Overall,the research of this thesis has two contributions.Firstly,the research of this thesis is helpful to the combination of futures market,traditional quantitative strategy and machine learning model.Secondly,the model constructed in this thesis has a certain reference value for the follow-up research of futures direction volatility prediction.