Research And Application Of Stock Index Futures Price Forecast Based On Neural Network

Since the birth of the financial market,investors have sought various ways of forecasting prices to increase investment returns.Current research methods include technical analysis,time series analysis,and machine learning methods.Neural networks have the characteristics of self-learning,high abstraction,and strong robustness.Using neural networks to predict prices is becoming more and more frequent.The traditional forecasting method has low forecasting accuracy;the forecasting results have poor applicability in financial engineering.This paper aims at the above two issues and makes the following three researches and applications.First,this paper designs a recurrent neural network that predicts high accuracy.Experimental results analysis of prediction error ratios and prediction accuracy of RNN,LSTM,and GRU.The results show that the GRU neural network with 8 neurons in the hidden layer has the best prediction performance.Secondly,this paper proposes a Trading-Position strategy.The trading strategy obtains trading signals based on the stock index futures information issued by the exchange.The strategy issued a bullish signal when holding sell orders decreased and buy orders increased;the strategy issued a short signal when the holding of sell orders increased and buy orders decreased.The experimental results prove that the strategy has a high annualized rate of return.Finally,this paper proposes a strategy based on neural network optimization for trading positions,uses GRU neural network to forecast prices,and then combines the signal of Trading-Position strategy,and finally obtains the Trading-Position based on neural network optimization.The experimental results show that the Trading-Position strategy based on neural network optimization is more profitable than the Trading-Position strategy.