Paper
CMG: A Causality-enhanced Multi-view Graph Model for Stock Trend Prediction
Published Oct 21, 2024 · Xi Cheng, Liang Wang, Yunan Zeng
Proceedings of the 33rd ACM International Conference on Information and Knowledge Management
0
Citations
0
Influential Citations
Abstract
The stock trend prediction problem refers to forecasting future stock price trends. In recent years, some methods discovered causal relations between stocks to address this problem. However, traditional causal discovery methods face unique challenges in the stock market, as they fail to uncover accurate causal relationships when a distribution shift happens in stock. Additionally, current methods also overlook the commonalities and differences between stock relations. To address these shortcomings, we propose a causal-enhanced multi-view temporal graph model, named CMG. This method explores comprehensive causal relations by incorporating distribution shift confounder and constructs a multi-view contrastive learning module to unearth the commonalities and differences between stock relations, thereby enabling more accurate stock trend predictions. Further experimental results and investment simulations demonstrate the effectiveness and profitability of CMG.
The CMG model accurately predicts stock trends by exploring causal relations and identifying commonalities and differences between stock relations.
Full text analysis coming soon...