计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (3): 308-314.DOI: 10.3778/j.issn.1002-8331.2008-0337

• 工程与应用 • 上一篇    

改进的多层图注意力网络在股价预测中的应用

姜明华,陈赟   

  1. 武汉纺织大学 数学与计算机学院,武汉 430200
  • 出版日期:2022-02-01 发布日期:2022-01-28

Prediction of Stock on Improved Hierarchical Graph Attention Network

JIANG Minghua, CHEN Yun   

  1. School of Mathematics and Computer Science, Wuhan Textile University, Wuhan 430200, China
  • Online:2022-02-01 Published:2022-01-28

摘要: 利用关系数据进行股价预测的方法最近已经被提出,但目前还没有找到一种有效的方法可以有选择地聚合不同类型的关系数据去预测股价。提出一种改进的多层节点图注意力网络(FHAN)模型,该方法融合Fraudar算法,提供了一种对多个对象关系之间看问题的视角。模型把公司看做节点,把交互看成边,选择性地聚合不同关系类型的信息,并将这些信息添加到每个公司的节点表示中,添加了信息的节点表示被输入到特定任务层自动选择信息,实验结果表明,该方法比目前流行的神经网络算法在股价预测的效果上更准确,实验选取不同神经网络算法做对比,在最优参数条件下,采用该方法比现有方法准确率平均提高约4%,最高提高约24%。


关键词: 股价预测, 多层节点图注意力网络, 关系数据

Abstract: Methods of useing relational data for stock market prediction have been recently proposed. No existing work on stock market prediction has an effective way to selectively aggregate information on different relation types. To address this, FHAN model include Fraudar method, provides a perspective to see the relationship between multiple objects. The model regards the company as a node and the interaction as an edge, which selectively aggregates information on different relation types and adds the information to the representations of each company. Node representations with the added information are fed into a task-specific layer. FHAN which can automatically select information outperformed all the existing main methods.

Key words: stock market prediction, multilayer node-graph attention network(FHAN), relational data