LSTM Neural Network's Application In Prediction Of Stock Price

The current development of computer technology has promoted the study of the stock market into a new stage.Neural network machine learning algorithms are widely used in the prediction of financial time series.Its good nonlinear approximation ability and self-learning adaptive characteristics make it one of the most popular forecasting methods.LSTM neural network is a new recursive neural network model that can be realized.Its characteristics of selective memory and internal influence of time series are very suitable for the prediction of stock price time series.Based on the analysis of the prediction of the stock price and the comparison of various neural network prediction methods,this paper discusses the feasibility of the short-term trend forecasting of stock price using the LSTM neural network optimized by Adam algorithm.The research data specifically selected for the paper is the data of the three stock indexes and eighteen active stocks in China stock market from August 2,2010 to May 31,2018.The data includes the date,opening price,closing price,the lowest price,the highest price and the daily volume.According to theoretical research and comparative verification,the LSTM neural network model can learn the changes in the historical data of the stock market,find out the influence and relationship between time series,and can use the advanced memory learning function of selective memory to dig deep into the time series of stocks.The results show that although LSTM neural network has a certain degree of time lag but its measurement accuracy is stable at a relatively high level,and has excellent trend prediction ability.