- The World Bank has developed an AI-powered model that analyzes over 90 variables to forecast refugee arrivals 4 to 6 months in advance, enabling governments to shift from reactive to anticipatory responses.
- This model facilitates early action, strengthens host-refugee relations, and enhances public service delivery.
- It is also adaptable for predicting and explaining other development challenges, such as poverty levels and macro-fiscal pressures.
Why this tool?
Every year, millions flee their homes due to conflict, climate change, or other crises. When refugees arrive in large numbers, host communities struggle as basic public services—health, education, and water—are overstretched.
Resources often arrive too late or not where they’re needed most. To provide a better response, the World Bank has developed an AI-powered model to support the Displacement Crisis Response Mechanism, or DCRM—the world’s first displacement risk financing mechanism.
How does it work?
The AI model uses machine learning to predict refugee arrivals 4–6 months in advance. It analyzes over 90 variables, like conflict, economic shifts, climate data, and even news and social media language, to anticipate movement.
This prediction data enables governments to build water points, expand health facilities, and construct classrooms before refugees arrive.
Case Study: Anticipatory Capacity for Uganda’s Displacement Crisis Response Mechanism (DCRM)
The model has been tested in Uganda, measuring UNHCR’s daily refugee arrivals data from 2014 to 2023. Data scientists collected the data for refugees coming from South Sudan and the DRC, producing a model with high predictive accuracy. It identified key drivers such as conflict, economic activity, climate, food prices, and language about gender, conflict and government.
When refugee inflows increased pressure on Ugandan services, the DCRM was triggered and funds were disbursed to scale-up service capacity – water access, health center capacity, and classroom expansion. By increasing public service availability, we can reduce tension between refugee and host communities and strengthen resilience.
Model Customization within the World Bank
This model can be used by the World Bank to identify which factors drive change in refugee flows. This evidence helps better target operations and manage displacement.
In the future, the model could also be adapted to predict and explain other important development issues, such as poverty levels or macro-fiscal pressures.