A Hybrid ABM-ML Framework for Simulating Climate Induced Migration in Burkina Faso
- 2026 Second International Conference on Multi-Agent Systems for Collaborative Intelligence (ICMSCI) : 320-326
Résumé
The internal migration, influenced by intricate socioeconomic and climatic factors, presents a considerable obstacle to sustainable development in Burkina Faso. Although the factors influencing migration are generally comprehended, quantitative, long-term forecasts of migratory trends are lim-ited, obstructing proactive policy formulation. This paper presents a hybrid modeling framework that integrates an Agent-Based Model (ABM) with a Machine Learning (ML) model to simulate interprovincial migration patterns up to the year 2100. The framework employs a Random Forest (RF) classifier, trained on historical data, to forecast household-level migration decisions within a spatially explicit ABM executed on the GAMA plat-form. We model migration dynamics using three Shared Socio-economic Pathways (SSP1, SSP2, and SSP5). Results show that Kadiogo, the capital province, is a major source of people leaving the country, while Boulkiemdé, Comoé and Houet, two rural provinces, are the main places people go to. The SSP2 scenario shows that migration flows will speed up a lot after 2060, which shows how demographic pressures will change in the future.
Mots-clés
climate-induced migration, internal migration, interprovincial migration, agent-based modeling, random forest, Shared Socioeconomic Pathways, migration forecasting, household migration decisions, spatially explicit simulation, GAMA platform, Burkina Faso