Summary
ABSTRACT
In this research article, Amundi Quantitative Research explores the use of graph theory and neural networks in asset management. In particular, they show how new alternative data such as supply chain databases require new tools to fully exploit this information. After reviewing the basic tools of graph learning theory, especially graph convolutional and attention networks, they compare the performance of some equity strategies when incorporating sector-based, correlation-based and supply-chain-based graphs. In particular, they show that supply-chain-based graphs are more and more informative these last years. This research opens the door to many applications of graph neural networks in asset management, e.g. ESG networks, ownership relations, infrastructure dependencies, physical risk, green energy markets.
To find out more, download the full paper
Authors
