Year: 2025
Type: conference-paper
Abstract: To better cope with the volatility of stock prices, this paper proposes an innovative HDA-LSTM (Hierarchical Double Attention-Long short-term model) model, which combines the long and short-term memory network (LSTM) with the attention mechanism and dynamically allocates the attention mechanisms by building a two-layer attention mechanism layer, and dynamically allocates the weights, thereby improving the accuracy and robustness of stock price prediction. In this paper, a dataset of stocks from the Shanghai Stock Exchange (SSE) Small and Medium Index Board (SMB) is selected to predict the closing prices after several days. A comprehensive comparison is made with the traditional machine learning model, and the single-layer LSTM-attention mechanism model. The results indicate that the HDA-LSTM model performs excellently, demonstrating strong generalization and adaptability in handling complex time series analysis. This suggests that the model provides an effective and reliable approach to stock market forecasting.
Author: Siyuan Yao
Institution: China University of Mining and Technology - Beijing
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W4412789753