Stock Price Prediction Using CNN-BiLSTM-Attention Model

Accurate stock price prediction has an important role in investment. Because data are characterized by high frequency, nonlinearity, and long memory, predicting prices precisely is challenging. Various forecasting methods have been proposed, from classical time series to machine-learning-based methods, such as random forest (RF), recurrent neural network (RNN), convolutional (CNN), Long Short-Term Memory (LSTM) networks their variants, etc. Each method can reach a certain level of accuracy but also its limitations. In this paper, CNN-BiLSTM-Attention-based model proposed boost the indices. First, temporal features sequence extracted using (CNN) bi-directional short-term memory (BiLSTM) network. Then, attention mechanism introduced fit weight assignments information automatically; finally, final results output through dense layer. The was first used predict Chinese index—the CSI300 index found be more accurate than any other three methods—LSTM, CNN-LSTM, CNN-LSTM-Attention. order investigate whether robustly effective indices, indices China eight international were selected test, robust effectiveness CNN-BiLSTM-Attention confirmed. Comparing with LSTM, CNN-LSTM-Attention models, it that highest almost all cases.