Embedded Files
Financial markets generate large volumes of heterogeneous and highly interconnected data. Prices, volatilities, firm characteristics, news, and economic relationships cannot always be adequately represented by standard tabular models, since dependencies between assets often play a central role in market dynamics.
My research in this area explores how machine-learning methods, and graph-based learning in particular, can exploit these complex dependence structures. The objective is to combine information on individual financial assets with the network of relationships linking them, whether these relationships arise from market co-movements, economic connections, or textual information. Applications include multivariate volatility forecasting and stock-movement prediction using graph neural networks and multimodal data.
A central challenge is to develop models that improve predictive performance while remaining robust to the non-stationarity, high dimensionality, and strong dependence that characterize financial data. More broadly, this research investigates how modern machine-learning architectures can complement traditional econometric approaches for forecasting and decision-making in finance.
Chen, Q. and Robert, C. (2022). “Multivariate Realized Volatility Forecasting with Graph Neural Network.”
Proceedings of the 3rd ACM International Conference on AI in Finance (ICAIF ’22), pp. 156–164.
Abstract
Financial economics and econometrics literature demonstrate that the limit order book data is useful in predicting short-term volatility in stock markets. In this paper, we are interested in forecasting short-term realized volatility in a multivariate approach based on limit order book data and relational stock market networks. To achieve this goal, we introduce Graph Transformer Network for Volatility Forecasting. The model allows combining limit order book features and a large number of temporal and cross-sectional relations from different sources. Through experiments based on about 500 stocks from S&P 500 index, we find a better performance for our model than for other benchmarks.
Chen, Q. and Robert, C. (2022). “Graph-Based Learning for Stock Movement Prediction with Textual and Relational Data.”
The Journal of Financial Data Science, 4(4), p. 152
Abstract
Predicting stock prices from textual information is a challenging task due to the uncertainty of the market and the difficulty in understanding the natural language from a machine’s perspective. Previous research has mostly focused on sentiment extraction based on news associated with a single stock; however, the stocks on the financial market can be highly correlated, as news regarding one stock can quickly impact the prices of other stocks. To take this effect into account, the authors propose a new stock movement prediction framework: multigraph recurrent network for stock forecasting (MGRN). This architecture allows combining the textual sentiment from financial news and multiple relational information extracted from other types of financial data. Through an accuracy test and a trading simulation on the stocks of the STOXX Europe 600 index, the authors demonstrate a better performance of their model compared with other benchmarks.
Google Sites
Report abuse