Statistics Netherlands Remote Access Environment (RA) and the ODISSEI Secure Super Computer (OSSC) provide an opportunity to analyse complex and highly sensitive datasets across various socio-economic domains. These datasets are increasingly analysed as what computer scientists call a graph and social scientists typically call a network. Analysing data as a graph opens up a wide range of analytical possibilities, including societal dynamics and previously underexplored connections between socio-economic domains. However, to fully harness the potential of this graph-structured data—and to fully exploit the competitive advantage that the Netherlands has within these resources, accessible tools that serve both Social Science and Computer Science communities are needed.

What is NetAudit?

NetAudit is a project within ODISSEI that is led by Dr. Megha Khosla at TU Delft’s Intelligent Systems Department. NetAudit bridges the gap between social and computer scientists by developing tools to analyse large population networks through interpretable network embeddings. Network embeddings are a way to convert the complex structure of a graph into a simpler format, typically a set of numbers, that computers can more easily work with. Imagine a social network where each person is a ‘node’, and the connections (friendships) between them are ‘edges’. Graph embeddings take this web of relationships and turn it into a list of numbers for each person, while keeping important information like who they’re connected to and how similar or influential they are in the network. This allows machine learning models to use these numbers for tasks like predicting new connections, classifying nodes, or understanding patterns in the graph.

NetAudit maximizes the potential of these embeddings by adding meaningful interpretations to each of their dimensions. These interpretations or explanations, rooted in the graph’s structure, allow domain experts to explore and uncover valuable insights—effectively helping them “find a needle in a haystack”. By enhancing the interpretability of this data, NetAudit will make it easier to apply advanced machine learning techniques for socially relevant tasks, ultimately unlocking new insights and fostering opportunities for interdisciplinary research. Once completed, the project will make these network embeddings available to the research community generally in a form that is well documented and interpretable so as to support onward usage in a wide range of research projects. 

Picture by Planet Volumes on Unsplash