Research On Futures Price Prediction And Quantitative Strategy Based On Deep Learning Algorithm

In recent years,with the rapid development of data science and technology,algorithm technologies such as machine learning,deep learning and reinforcement learning are increasingly frequently applied in the field of financial time series,and the prediction of financial time series has become a research hotspot.At the same time,financial quantitative investment strategy is also developing rapidly.This thesis will build an international gold futures price prediction model based on machine learning and deep learning algorithms,formulate appropriate quantitative trading strategies for gold futures according to the results of the model prediction.Finally,back test will be conducted according to the prediction results and quantitative trading strategies to explore the yield of the model and quantitative strategy.Firstly,the original sample data was cleaned and sorted to obtain relatively neat sample data.Then,according to the sample data,multiple technical indicators of gold futures were constructed.The feature screening of the sample data set was carried out using Lasso regression model,and finally the usable sample data set of international gold futures was obtained.According to the obtained sample data,a base model of international gold futures price prediction based on machine learning and deep learning algorithm was constructed.The prediction ability of 10 base models was compared according to the two evaluation indexes of root mean square error(RMSE)and R~2,and the base model with good prediction effect was integrated to obtain an international futures price prediction model based on ensemble learning.In the quantitative timing part of this thesis,we set trading signals,take the maximum retracement ratio as the evaluation index,and use the method of machine learning to quantify timing.Based on the international gold futures price prediction model and quantitative trading strategy constructed above,under the condition of not considering the margin,assuming that the initial capital is 1 million and the position ratio is full,all parameters of the trading strategy are set.The closing price of the daily frequency data of gold futures from January 1,2018 to January1,2019 in the sample data set was selected to carry out the quantitative timing strategy backtest,and the yield of the forecast model was 1.069%,and the yield and win rate were 10693.46.Under this quantitative investment strategy,the yield and win rate performed well,indicating the performance of this quantitative trading strategy was good.