Values are based on migration estimates generated using a deep learning model proposed by Gaskin and Abel (2025). The framework combines and reconciles multiple data sources, including UN DESA migrant stocks, bias-adjusted flow estimates derived from Facebook data, harmonised statistics from European countries via QuantMig, and official data from select nations. To improve estimation accuracy—particularly in data-poor countries with no prior migration data or official statistics—the neural network leverages a wide range of geographic, economic, cultural, and political covariates as predictive variables.
To compare these annual results with longer-term five-year migration flow estimates based on demographic accounting models, visit the original Global Migration Flows Explorer.
Zenodo Download — HuggingFace Download
Gaskin, T., & Abel, G. J. (2026). Deep learning four decades of human migration. Nature 655, 148–157 (2026). Advance online publication. doi.org/10.1038/s41586-026-10611-7