A Study On The Improvement Of LSTM Algorithm And Its Application In Stock Index Prediction

With the continuous development of China’s economy,stock investment has become a popular form of investment and financial management in China.The movement of stock prices can directly affect not only the stability of the stock market,but also the healthy development of the economy and finance.Investors often need to benefit from stock price movements by predicting them,while governments and other relevant agencies need to regulate and manage the market in a timely and effective manner.However,as a highly complex and dynamic system,the stock market is influenced by many factors,including internal factors such as business trends and external factors such as macroeconomic trends.The complexity and unpredictability of stock price movements are due to the volatility,non-linearity and low signal-to-noise ratio of the stock market,among other factors.The changes in the stock market can be reflected in a timely and comprehensive manner by the stock index,which can reflect the changes in stock prices.Therefore,it is necessary to effectively predict the stock index.In order to make effective forecasts of financial stock indices,this paper proposes an adaptive fusion stock index forecasting model based on Bayesian optimization: the model is an integrated model consisting of a combined forecasting model of empirical modal decomposition(EMD)and long and short-term memory(LSTM)models and a single LSTM model,which uses stock index volatility to determine the applicable situation and thus adaptively selects the optimal model,and the EMD-LSTM and The hyperparameters of the LSTM model are automatically optimized using a Bayesian optimization algorithm.In this paper,the statistical description of stock indices is firstly conducted and the volatility of the three stock indices is found to be significantly different.Modeling the data with respect to this feature,comparing the model proposed in this paper with traditional time series models and machine learning models for a total of six models,the comparative study finds thatFrom the prediction results of Chinese stock indices: EMD-LSTM model has better results in predicting the closing price of CSI 300 index and SZSI closing price,and LSTM model has better results in predicting the closing price of SSE index.From the results of the U.S.stock index prediction: EMD-LSTM model has good effect in predicting the closing prices of Dow Jones Index,S&P 500 Index and Nasdaq Index,which verifies the generalizability of the model in this paper.From the above empirical analysis,it is concluded that the model can predict the stock indexes according to the volatility of the data,and the model can effectively predict the stock indexes with different volatility.Therefore,the model is useful as a guide for actual investment.