Filling survey gaps in food security monitoring with spatio-temporal additive Gaussian process models

This repository implements a Bayesian spatio-temporal additive Gaussian process model for forecasting food insecurity (Food Consumption Score) at the sub-national level in Nigeria and Chad, benchmarked against a Bayesian Ridge with BYM2 spatial effect, a Multi-Layer Perceptron, and XGBoost. The GP model is estimated via Stan/CmdStanR, with R scripts for data preparation, cross-validation, and diagnostics, and Python scripts for the MLP, XGBoost. Model outputs, figures, and metric tables used in the paper are included under output/. The underlying WFP survey data are not included and are available upon reasonable request.