A Research On Prediction Of Stock Price Based On LSTM

As we all know,stocks are one of the important components of our country’s economy.In recent years,with the continuous improvement of people’s living standards,more and more people are carrying out stock investment transactions.Therefore,it is very important for investors to predict the stock price more accurately.Therefore,more and more investors are beginning to study stock predictions.General investors use traditional technical analysis methods to select stocks for trading,while senior technical analysts combine basic indicator data and technical indicator data to predict stock prices.In order to further study the stock price trend,researchers use mathematical model analysis to make predictions based on the historical data of the stock.With the development of neural networks,especially the rapid development of deep learning,deep learning has achieved good results in time series prediction.As one of the most classic models of recurrent neural networks,LSTM has received extensive attention.In recent years,using the neural network methods to predict stock data(such as stock finance and stock data)has become the focus of research.With the development of quantitative investment concepts and algorithmic trading concepts,more and more people are using neural network systems to predict stock prices and stock trends.However,so far,we still don’t have good guidelines to build hidden layers of neural networks.Therefore,many researchers still adjust the parameters based on the experience explored by predecessors or based on their own experience.This article uses the method of stock price prediction based on LSTM model.The data of this article comes from VTI stock data of Yahoo Finance,During the experiment,several relevant characteristic indicators that affect the stock closing price were introduced into the model.At the same time,hidden layers based on Dropout and dense were used to build a LSTM model,and the hyperparameter information of this model was continuously adjusted.Then we obtain the final LSTM model,and predict the adjusted closing price of the stock based on the obtained LSTM model,and compare it with other basic models or algorithms.Experimental results show that,compared with the traditional Last Value model mentioned in this article,the accuracy of the LSTM prediction model is improved by 8%;compared with the classic XGboost model,the accuracy is also improved by 3%.Compared with traditional prediction methods,the LSTM model performs well on the results of RMSE,R2,and error values;therefore,using the LSTM model based on relevant characteristic indicators is a good choice for stock price prediction.