Research On Stock Price Prediction Based On LSTM Model And Multi-source Data Fusion

Stocks can promote the flow of capital in the market and promote the development of China's economic,so the study of stock price fluctuations has always been a hot research issue.Traditional stock research generally uses its historical transaction data to build traditional time series models for stock price forecasts.With the development of science and technology,there is a lot of new information emerging in financial markets that affects stock price volatility.Therefore,only using stock time series data to predict stock prices has certain limitations.This paper will start from the deep learning model and use multi-source stock information to build a multi-source stock data fusion model,so as to improve the accuracy of stock price prediction.This paper mainly discusses the LSTM model of stock historical transaction data and industry data fusion.Firstly,the random forest algorithm is used for feature selection of stock indicators,and it is found that the closing price has the highest score.Secondly,the comprehensive factor of each industry is extracted as the industry index through factor analysis.Then,the DTW algorithm is proposed to measure the correlation between the target stocks and various industries.Finally,three fusion schemes and a control group are proposed,and the control group has only the target for individual stocks,the first fusion scheme is the integration of the target stock with its industry,the second fusion scheme is the fusion of the target stock with the DTW industry coefficient adjustment module,and the third fusion scheme is the fusion of the target stock and its most relevant industry.This paper uses the stock data from 2020 to 2021 to conduct experiments,and the conclusion shows that the third fusion scheme can improve the accuracy of stock price prediction to a certain extent.