Reproduce “Deep learning four decades of human migration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Thomas Gaskin, Guy J. Abel
Gaskin and Abel developed a deep recurrent neural network combined with a mechanistic flow model to generate the first comprehensive, annual dataset of origin-destination human migration flows and stocks between 230 countries from 1990 to 2023. The model, trained on diverse covariates and existing migration data, provides robust estimates with quantified uncertainty, outperforming traditional methods in temporal resolution and accuracy.
Key insights
- Deep recurrent neural networks can effectively learn complex, non-linear relationships and long-range temporal correlations in human migration patterns, moving beyond the Markovian assumption of traditional models.
- Integrating diverse covariate information allows the model to disentangle underlying drivers of migration, revealing varying sensitivities to factors like life expectancy, GDP per capita, and religious similarity.
- Quantifying uncertainty in migration estimates highlights regions with inconsistent data, providing critical guidance for future data collection efforts and improving the reliability of analyses.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.