Application Of Integrated Learning Based On ARMA Model And GRU Model In Quantitiative Investment

To be able to predict the occurrence of an event has always been a problem that people want to discuss.Through careful and accurate prediction,we can make some preparations ahead of time,and if something better happens,we are actively prepared to make it better.If you predict something bad is about to happen,be prepared to minimize the loss.With the continuous progress of science and technology,human society and other aspects,people more and more want to predict the development of future events according to the present situation.Therefore,the time series prediction method has been developed by leaps and bounds.At present,there are two main methods of prediction.The first is a classical time series prediction method,such as ARMA et al.Due to the long-term development,time series has been widely used in various industries,but this method requires a lot of data,and sometimes requires a lot of assumptions,so practical applications often waste a lot of time.But sometimes we can't get a satisfactory result,especially in the face of the long-term trend of the stock market.The second is the centralized prediction method in machine learning,the most typical one is the neural network model.Because the neural network is not confined to assumptions and has a wide range of applications,it has been paid great attention since the beginning of its generation.But many of its models are too complex to lead to transportation.The calculation time is too long even produces the over-fitting phenomenon.In the stock market,with the passage of time,listed companies increase,domestic economy and foreign countries more closely linked,more and more data,more and more complex,the difficulty of forecasting also increased,appeared the concept of quantitative investment.For this reason,there are many scholars trying to combine them to avoid their shortcomings while ensuring their advantages,and then improve the accuracy of prediction in the use of quantitative investment.In this paper,we first combine the autoregressive model(AR model)in time series with the threshold regression model(GRU model)in machine learning to get the AR-GRU model.It is used to study the daily data of Shanghai Stock Index from January 2005 to July 2018,and then forecast the data according to the learning results,and analyze the deviation of the data by comparing the real results and the forecast results.Another kind of ensemble learning model is proposed in this paper,which integrates the moving average model(MA model)and threshold regression model(GRU model)in time series to get the GRU-MA model.It is also used to study the daily data of Shanghai Stock Index from January 2005 to July 2018.Then the data are predicted according to the learning results,and the deviation is analyzed by comparing the real results and the forecast results.Finally,the prediction results of AR-GRU model,GRU-MA model,ARMA model and GRU model are compared,and the prediction accuracy of the new model is higher than that of the original model.