Research On Quantitative Timing Strategy Based On Bayesian-Adam-LSTM Neural Network

With the continuous development of China’s securities market and the arrival of the big data era,quantitative investment has quickly entered the public eye with its powerful information processing ability and the advantage of eliminating subjective emotional interference.At the same time,deep learning is also widely applied in various fields due to its excellent generalization ability,high accuracy,and efficiency.The research topic of whether combining deep learning with quantitative investment can achieve better results is receiving increasing attention.This paper selects the daily data of the Shanghai and Shenzhen 300 stock indexes from January 1,2013 to December 31,2022 as the experimental data,combines the deep learning model with the optimization algorithm,builds a short-term memory neural network model(Bayesian Adam LSTM)combining the Bayesian method and Adam optimization algorithm to predict the stock price,and constructs timing strategies based on the model prediction results,which performs well in the return period.The main research work and conclusions of this article include:(1)Selecting four types of technical indicators,including trend,overbought,oversold,trading volume,and energy,totaling 18 candidate factors,and using basic market data to calculate the indicator values;(2)In order to solve the multicollinearity problem,principal component analysis(PCA)is used to reduce the dimensions of indicators,and five principal components covering the original data information are obtained;(3)The Bayesian Adam LSTM prediction model is constructed with five principal components as input variables and closing price as output variables.The model is based on multi-layer LSTM,and uses Adam optimization algorithm and Bayesian optimization algorithm to optimize hyperparameter;(4)Based on the predicted results and the double moving average strategy as the theoretical basis,a trend timing strategy was constructed,and a real market backtesting was conducted from January 1,2021 to December 31,2022.The results showed that the quantitative timing strategy based on the Bayesian Adam LSTM neural network can outperform the benchmark returns and has excellent performance in terms of profitability and risk level.Compared with existing literature,the main innovation of this article is to combine technical indicators with the Bayesian-Adam-LSTM neural network model for stock price prediction,and construct a quantitative timing strategy based on the prediction results.The research conclusion can provide a certain timing reference for institutional and individual investors.