Stock Index Prediction Based on CNN-BiLSTM-Attention
Yuxuan Yang, Jingchuan Xu · 2024
This paper proposes a new model CNN-BiLSTM-AM in the field of stock index prediction. This model analyzes important market indicators such as opening price index, highest price index, and trading volume, and combines the advantages of the three models to predict the complex and ever-changing stock index market. The study conducted model training and testing based on the Shanghai Composite Index and Shenzhen Component Index data from 2012 to 2024. Experimental results show that on multiple evaluation indicators, including mean square error (MAE), mean square error (MSE), root mean square error (RMSE), coefficient of determination (R2), relative percentage deviation (RPD) etc., this model is significantly better than existing methods and shows high prediction accuracy. This article not only provides a new perspective in the field of financial forecasting, but also lays the foundation for future applications in different markets and time scales. Future research will be dedicated to further optimizing this model to improve its performance and expand its application range.