Graph Neural Networks in Financial Markets: Modeling Volatility and Assessing Value-at-Risk

Abstract

This paper explores the application of graph neural networks (GNN) in the prediction of volatility and value-at-risk (VaR) calculation in financial markets, aiming to provide a new method for financial risk management. With the acceleration of globalization, financial markets have become more complex and interconnected. Traditional statistical models make it difficult to handle a large number of nonlinear relationships, high-dimensional data, and dynamic changes. In recent years, the development of machine learning, especially deep learning technology, has made graph neural networks a model that can effectively capture the interactive relationship between nodes in complex systems and is widely used in many fields. The financial market can be regarded as a network structure composed of multiple assets. Each asset is a node in the network, and the correlation between their price changes constitutes an edge. Therefore, using graph neural networks to model financial markets has a natural advantage. It can not only consider the historical performance of a single asset but also improve prediction accuracy by learning the mutual influence between different assets. The experimental results show that compared with other deep learning models including BiLSTM and CNN, GNN performs well in evaluation indicators such as MSE, RMSE, and MAE, showing its superiority in the field of financial volatility prediction. This research not only provides the financial industry with a new way to understand and manage risk, especially in a highly uncertain and rapidly changing market environment, but it is also expected, with the accumulation of more high-quality data sets and technological advances, to play a more important role in the future.