A NOVEL MODEL FOR STOCK CLOSING PRICE PREDICTION USING CNN-ATTENTION-GRU-ATTENTION

Predicting stock price to avoid risk is the focus of stock research. A reliable predicting model could offer insights in stock price fluctuations and ultimately could provide the opportunity of gaining significant profits. In this paper, a new composite forecasting model is proposed to forecast the stock closing price of the next trading day. This model consists of three parts. Convolutional Neural Network (CNN) is used to collect: the factors that affect the stock price. attention mechanism (Attention) is used to compute the impact of stock data at different times on stock price. Gate Recurrent Unit (GRU) is used to forecast the stock price. It can make good time series prediction. Through comparison with CNN-Attention-LSTM-Attention, CNN-Attention-GRU, CNN-GRU-Attention and other traditional models. The experimental results indicate that the performance of this model is better to other models, and it has the best performance in evaluation metrics like MAE, RMSE and R-2. It is more appropriate for stock price prediction.