Nicolai Bloch Jessen
Explainable dynamic graph neural networks in an emerging market bond portfolio context
Due to the increased use of artificial intelligence (AI) and machine learning (ML) the demand for explaining these often called “black box” methods have increased as well. The use of graph neural networks (GNN) has proven state-of-the-art performance in recent research, but the research on dynamic GNN is limited, and there still exists a large research gap in explaining the predictions of a dynamic GNN. Therefore, the main objective of this industrial PhD project is to propose a method for explaining the predictions of a dynamic GNN.
As reinforcement learning (RL) have proven superior performance in portfolio optimization tasks, the second objective of the project is to make a RL optimized portfolio of emerging market sovereign hard currency bonds, that outperforms a relevant benchmark in terms of risk adjusted returns. The results of this project will be very important for the financial sector, but it could also be of great importance for all industries where explainable time series models are valuable.