NetAudit
Making Dutch population-network embeddings more interpretable for research
Duration: April 2024 - May 2025
Funding: NWA ODISSEI Roadmap grant, Task 4.4
NetAudit focuses on building understandable network embeddings for the Dutch population network.
The project sits at the intersection of social science and machine learning. Its goal is to turn very large network data into representations that remain useful for analysis while being easier for researchers to interpret.
By giving embedding dimensions clearer meaning, NetAudit supports exploratory analysis, downstream prediction tasks, and broader reuse of population-scale network data in interdisciplinary research.
The main outcome of the project is a population-scale study of the Dutch social network built from shared neighborhood, work, family, household, and school contexts. The resulting analysis shows that network embeddings capture meaningful structural differences, and that an interpretable dimension tied to educational ties and attainment is associated with right-wing populist voting.
Another key deliverable of NetAudit is the release of both untransformed and transformed population network embeddings for 2020, 2021, and 2022 within the secure remote access environment of Statistics Netherlands through the Storage Facility, in collaboration with ODISSEI.
Main result and publication
- Lüken, Malte, Garcia-Bernardo, Javier, Deb, Sreeparna, Hafner, Flavio, and Khosla, Megha. Population-Scale Network Embeddings Expose Educational Divides in Network Structure Related to Right-Wing Populist Voting. arXiv preprint, 2025.
Team
- Megha Khosla (Lead, TU Delft)
- Malte Lüken (Netherlands eScience Center)
- Flavio Hafner (Netherlands eScience Center)
- Sreeparna Deb (TU Delft)