TCN-LSTM Stock Price Prediction Based on Sentiment Analysis
In view of the fact that traditional stock price prediction models do not consider the impact of market investor sentiment on stock prices and are difficult to handle stock price prediction tasks,a time series Deep Learning model BERT-TCN-LSTM that integrates sentiment features is proposed.Firstly,sentiment analysis is performed on investor comments crawled from stock bars,and the average value of daily sentiment is extracted as the input of the model.Secondly,the TCN-LSTM model constructed by the daily sentiment average and stock price data and technical indicators is trained.Finally,experiments are conducted on the data sets of CSI 300 and four individual stock data.The results show that compared with the Temporal Convolutional Network(TCN),LSTM and CNN-LSTM,the Mean Absolute Error(MAE)of BERT-TCN-LSTM on the CSI 300 data set is reduced by an average of 54%.The BERT-TCN-LSTM model can effectively improve the accuracy of stock price prediction.
time series predictionsentiment analysisTemporal Convolutional NetworksLong Short-Term Memory
国家科技期刊平台
NSTL
万方数据
