Scholarlink

In recent years, the intricate interplay of corporate entities within business conglomerates has garnered attention due to its profound implications for financial markets, risk management, and strategic decision-making. Leveraging reliable data sourced from a commercial bank, this paper introduces a novel approach, EquityNet, using the Relational Graph Attention Network (RGAT) model and deep learning techniques to uncover and predict equity relationships among corporations. A hierarchical attention structure enhances the effectiveness of the model by dynamically assigning attention scores to neighboring nodes based on different types of equity relationships. To capture nuanced investment patterns, EquityNet employs the innovative Graphlet Degree Vector (GDV), expanding the receptive field and improving structural pattern distinctions between nodes. Experimental results on real-world banking data demonstrate the model’s superiority over traditional methods. This study contributes to a pioneering exploration of equity relationships using advanced GNN methodologies, offering insights into complex interactions within corporate networks.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EquityNet: Unveiling Corporate Equity Relationships in Business Conglomerates Using Graph Neural Networks and GDV Features

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • B. Li,
  • Bingce Wang,
  • Lifeng Li,
  • Weiping Li,
  • Tong Mo

摘要

In recent years, the intricate interplay of corporate entities within business conglomerates has garnered attention due to its profound implications for financial markets, risk management, and strategic decision-making. Leveraging reliable data sourced from a commercial bank, this paper introduces a novel approach, EquityNet, using the Relational Graph Attention Network (RGAT) model and deep learning techniques to uncover and predict equity relationships among corporations. A hierarchical attention structure enhances the effectiveness of the model by dynamically assigning attention scores to neighboring nodes based on different types of equity relationships. To capture nuanced investment patterns, EquityNet employs the innovative Graphlet Degree Vector (GDV), expanding the receptive field and improving structural pattern distinctions between nodes. Experimental results on real-world banking data demonstrate the model’s superiority over traditional methods. This study contributes to a pioneering exploration of equity relationships using advanced GNN methodologies, offering insights into complex interactions within corporate networks.