Financial Forecasting based on LSTM and Text Emotional Features

In order to realize the prediction of financial time series, this paper first collects relevant text information on Sina Finance, and extracts and obtains the time series of financial sentiment index with three indexes through sentiment analysis of the text. At the same time, this paper collects the time series of nine financial indices formed by the Shanghai Stock Exchange Index (SSE), and combines the two as the research object. Considering the long-term delay of financial time series, this paper chooses long short term memory (LSTM) neural network to construct prediction model. After training and testing the model, the results show that after increasing the text sentiment index, The mean square error (MSE) of the prediction results of the LSTM prediction model decreased from 74.57 to 19.06, and the mean absolute error (MAE) decreased from 5.96 to 3.14, and with the cyclic neural network (RNN) and the particle swarm back propagation neural network (PSO-BP). The prediction results are more accurate than the prediction model. The error is small and the accuracy is improved.