| The widespread adoption of Internet has greatly accelerated the distribution of information and data.Financial news,which tracks the movements of financial markets,requires full mining and analysis of financial data in order to provide accurate update of the markets and potential stock price indication.This paper collects extensive financial data from stock market and descriptive information of financial news,establishing a deep learning network based on LSTM through the extraction of data features,in order to analyzes the impact of financial news on stock prices.Firstly,in the collection and processing of financial data and discriptive information in the stock market,Selenium data collection method is used to gather structured financial data in the stock market,and PLSA algorithm is used to extract financial news keywords and convert financial text into structured data;Secondly,with regard to data feature extraction,having retained the time features of the variables to be predicted,SOCNN is used to extract the features of the mixed time series stock market data by introducing the entry mechanism.LTP Chinese language segmentation method is used to extract the key messages of financial news;In the aspect of model construction,the LSTM network is used to train and construct the time series prediction model,and the principal component analysis method and the denoising self-coder are used to reduce the dimension of high-dimensional financial time series data.The selection and design of activation function,empirical risk function and optimizer are studied.Using the stock price prediction model,with the experimental financial data as test sample,the paper examines the stock price movement of listed companies.The prediction results show that the LSTM deep learning network,after feature extraction by DAE denoising self-encoder,has improved prediction accuracy.It shows that the time series prediction model based on deep learning studied in this paper has a satisfactory prediction effect on financial time series. |
