An Attention-Augmented BiLSTM-CNN Model for Stock Price Prediction

The inherent volatility of the stock market has challenged investors in seeking reliable predictive models. This paper presents a hybrid deep learning model that combines bidirectional long short-term memory (BiLSTM) and convolutional neural network (CNN) architectures to forecast daily closing prices of stocks. To strengthen this architecture's performance, a weight-based attention mechanism is used to significantly enhance the model's performance. The experimental results of this novel architecture demonstrate the promising potential of the proposed model in multiple datasets. This model outperforms traditional methods of stock prediction using LSTM with ARIMA, LSTM-CNN, and other LSTM-based models. The improved performance of the proposed hybrid model highlights its potential for practical application in financial markets, offering investors a robust tool to make trades in the stock market full of uncertainties.