EGHAN: Event Graph Enhanced Hypergraph Attention Network for Stock Movement Forecasting

The forecasting of stock movement has been extensively researched. Due to the market's randomness and dynamism, accurately forecasting stock price trends is a highly challenging task. Currently, most methods for forecasting do not take into account the effects brought by multiple financial factors and their corresponding background relationships. Additionally, relevant datasets are relatively limited. In this study, we propose EGHAN: Event Graph enhanced Hypergraph Attention Network for stock movement forecasting, which employs hypergraphs to model financial news, historical prices, along with their background relationships. Hypergraph attention mechanism is employed to update node representations. Additionally, the model integrates an event graph and utilizes the knowledge within it to enhance the representation of financial news. To evaluate our method, we construct a new stock dataset and conduct extensive experiments, demonstrating the superior performance of our method.