Research On SSE 50ETF Price Forecast And Option Strategy Based On GA-LSTM Model

The construction of directional timing strategies based on the pre-judgment of the underlying asset price trend is one of the main options trading strategies.With the rapid development of options in the domestic market,research on option investment strategies has broad market prospects.But for a long time,due to the randomness and non-stationary characteristics of financial time series data,the trend prediction of financial assets such as stocks and their derivatives has been a difficult problem in directional timing trading.The deep learning model long and short-term memory neural network(LSTM)can deal with the non-stationarity and complexity of time series,and is suitable for price trend forecasting and constructing corresponding strategies accordingly.Based on the medium and high frequency volume and price data of the SSE 50 ETF,this article uses real-number-encoded genetic algorithms to optimize the time window width and the number of neurons in the hidden layer in the construction of the LSTM model,and uses the optimized GA-LSTM model to predict the SSE 50 ETF closing price.Secondly,this article constructs SSE 50 ETF option directional timing trading signals based on the predicted prices,superimposes the volatility judgment index to determine the buying and selling methods of option transactions,constructs an option intraday directional trading strategy that considers volatility factors.Loss of value.Finally,this article divides the underlying option market into strong and weak signal markets and non-signal markets through different directional signals,adds combination strategy trading in weak and non-signal markets,and combines strategic risk management measures to help strategies in weaker and weaker trading signals.Get more benefits when there is no directional signal,and realize strategy optimization.The research results in this paper show that the LSTM model has a smaller MSE for the option underlying ETF price prediction error than the RNN cyclic neural network model,indicating that the LSTM model has a stronger advantage in the financial time series forecasting problem,and the use of genetic algorithms to optimize the hyperparameters of the LSTM model can be To a certain extent,train the optimal LSTM model to improve the prediction accuracy.The option directional intraday strategy constructed with GA-LSTM model to predict prices and combined with volatility judgment indicators has a good backtest result,which reflects the effectiveness of the model in practice.The option investment strategy optimized by adding portfolio strategies and risk management measures can increase the potential benefits of the strategy and guarantee limited risks by increasing open positions in a weak and non-directional signal market.