Research on stock return prediction model based on hypergraph attention network

• 1、School of Computer Science (National Pilot Software Engineering School),Beijing University of Posts and Telecommunications,Beijing 100876

Abstract：The stock market has long been one of the most important investment choices for individuals and institutions in pursuit of wealth. Meanwhile, the rapid development of computer science and technology and market economy in recent years is profoundly changing the face of the financial industry, and they not only bring a series of challenges but also a wide range of opportunities for the development of the financial industry. In this paper, we propose a stock return prediction model based on hypergraph attention network (HAN) , which consists of a time series extraction layer, a hypergraph attention layer and a fully connected embedding layer. In the temporal embedding layer, GRU and temporal attention layers are utilized to extract the temporal embedding of stocks. In the hypergraph attention layer, a graph attention network (GAT) is utilized to learn the interactions between stocks. In the prediction layer, the temporal embedding vectors are fused with the stock relationship embedding vectors and processed by the full connectivity layer to obtain the ranking of future stock returns. Finally the model is subjected to prediction experiments on various sectors of the Chinese A-share market, and the results show that it outperforms other benchmark models.

Keywords： stock prediction；hypergraph Learning；graph attention network