Attention-based BiLSTM model for stock price prediction

Abstract

Abstract: Stock price prediction is a hot issue in the field of quantitative finance. Investors hope to discover the objective laws of stock price fluctuations from historical data, optimize their investment strategies, and avoid risks to obtain better investment returns. With the development of deep learning technology, neural networks have shown good forecasting effects in task of time series data forecasting. Aiming at the temporal correlation of stock data, a bidirectional LSTM network stock price prediction model fused with attention mechanism is proposed. The experimental results show that the bidirectional LSTM network can effectively learn the correlation between data when performing the prediction task, and the attention module helps the model to better capture the key information in the stock data. Compared with other prediction networks, the model has higher prediction accuracy and lower prediction error, and achieves the best prediction performance on different datasets, which can provide help for stock price prediction.