Research On Price Forecast And Timing Strategy Of Stock Index Futures Based On WA-LSTM

The stock market is known as a barometer of a country's macro economy,and stock index futures prices have a price discovery function for stock indexes.Accurate predictions on them can not only provide a reference for policy making,but also earn substantial investment returns for investors.Therefore,stock index futures price prediction has always been an important research direction in the field of financial investment.Quantitative investment is a new investment analysis technology that has emerged in recent years.Since its inception,investors have tried to use it in stock index futures investment transactions.However,due to the large number of investors,uneven levels and asymmetry in information,stock index futures prices have the characteristics of non-linearity,non-stationary and high noise,which to a certain extent increases the diffculty of the use of quantitative investment techniques to predict stock index futures prices.This article combines wavelet analysis(WA)in the field of signal processing and Long Shortterm memory(LSTM),which has a good ability to analyze time series signals,to construct a stock index futures price prediction model.For the prediction model,the model's determination coefficient in the test set reached 0.9683,and the accuracy rate of classification reached 53.5%.Compared with the WA-ANN model,the WA-LSTM model performs better in all evaluation indicators.Compared with the LSTM model,the WA-LSTM model has better classification effect,while the regression fitting effect is slightly inferior.On this basis,this article constructs a WA-LSTM-based stock index futures timing strategy,which simulates back-testing the historical market quotations of the Shanghai and Shenzhen 300 stock index futures from 2018 to 2020,and from the perspectives of risk return and position trading analyzed the backtest results.At the same time,two comparative experiments of stock index futures timing strategy based on LSTM model and stock index futures timing strategy based on WA-ANN model are designed to verify the effect of the strategy.Finally,this article further optimizes the initial strategy by including technical analysis indicators in the factor pool.The research in this paper shows that:(1)Using wavelet analysis to denoise stock index futures price signals can improve data quality and improve strategic performance;(2)Compared with traditional neural network models,Long short-term memory networks are more suitable for financial time series data analysis;(3)The stock index futures timing strategy constructed by the WA-LSTM model can obtain better investment returns,and LSTM model has huge development potential;(4)On the basis of basic volume and price data,through the introduction of technology Index factors can provide gain information for predictive models and improve strategy effects.