Advanced Graph Machine Learning Methods for Complex Socio-Economic Systems

The work will focus on developing AI methods for analysing complex, structured data, with a focus on graphs. It will investigate methodological improvements to Graph Neural Networks (or related models), based on Network Science and structural properties of the input graphs, such as node roles, connectivity patterns, etc. The project will also address the generation of synthetic graph data from real-world networks, aiming to expand experimental opportunities while mitigating constraints related to accessibility, confidentiality, and reproducibility. The main domains will be economic and financial networks and other socioeconomic systems. The project will involve extracting structured data from these domains using computer-vision neural network models capable of analysing digitised documents, reports, historical archives, and handwritten documents. These tools will then be applied to the FIS3 WeInIt project sources, to create an intergenerational database of wealth in Italy.