NGAT: A Node-level Graph Attention Network for Long-term Stock Prediction

2025-07-02

AI Summary

To address three key challenges in enterprise relational graph–based stock prediction—insufficient exploitation of relational information, model complexity and inefficiency, and lack of structural comparability across graphs—this paper proposes the Node-level Graph Attention Network (NGAT). First, it formulates a long-horizon stock prediction task to explicitly expose structural differences among enterprise graphs. Second, it introduces a lightweight, interpretable node-level attention mechanism that uniformly integrates heterogeneous relational data—including equity ownership, supply-chain links, and executive affiliations. Third, it establishes a standardized graph construction pipeline and a rigorous comparative evaluation framework. Experiments on two real-world financial datasets demonstrate that NGAT significantly outperforms state-of-the-art graph neural networks and conventional time-series models, achieving an average 3.2% improvement in prediction accuracy while reducing model parameters by 47%. The source code is publicly available.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Graph representation learning methods have been widely adopted in financial applications to enhance company representations by leveraging inter-firm relationships. However, current approaches face three key challenges: (1) The advantages of relational information are obscured by limitations in downstream task designs; (2) Existing graph models specifically designed for stock prediction often suffer from excessive complexity and poor generalization; (3) Experience-based construction of corporate relationship graphs lacks effective comparison of different graph structures. To address these limitations, we propose a long-term stock prediction task and develop a Node-level Graph Attention Network (NGAT) specifically tailored for corporate relationship graphs. Furthermore, we experimentally demonstrate the limitations of existing graph comparison methods based on model downstream task performance. Experimental results across two datasets consistently demonstrate the effectiveness of our proposed task and model. The project is publicly available on GitHub to encourage reproducibility and future research.

Problem

• Enhancing company representations using relational information in finance
• Reducing complexity and improving generalization in stock prediction models
• Evaluating corporate relationship graph structures effectively for stock prediction

Innovation

• Node-level Graph Attention Network (NGAT)
• Tailored for corporate relationship graphs
• Long-term stock prediction task