Predictive Displacement for Anticipatory Action (PREDICT)

The surge in forcibly displaced persons globally is outpacing humanitarian responses, exacerbated by unfolding crisis. This diminishes the capacity of aid organizations to provide timely support, leaving vulnerable populations without necessary services for extended periods. Consequently, timely preparedness is vital for securing organizational and financial resources, enabling swift and effective interventions wherever required.

With global displacement and humanitarian needs rising, PREDICT leverages a machine learning developed by the Danish Refugee Council (DRC) to forecast displacement 3-4 months in advance. Using open-source data on conflict, health, environment, food insecurity and income, the Anticipatory Humanitarian Action for Displacement (AHEAD) model developed by PREDICT helps WFP anticipate crises and take early action.

The sprint aims to strengthen WFP's early warning system in its capacity to anticipate spikes in forced displacements and detect where humanitarian needs and food insecurity are likely to significantly increase. The model is currently being tested in DRC, Ethiopia and Sudan.

PREDICT improves displacement early warning, further enhancing WFP’s ability to detect future humanitarian needs and food insecurity. The team has published AHEAD PREDICT Dashboard which showcases displacement trends within a 3-month forecasting window. The data is now integrated into WFP to support programme planning.

Data from a pilot run by the Danish Refugee Council shows teams on the ground have been able to halve humanitarian response time in Burkina Faso, from 80 to 40 days on average. Other pilot data from South Sudan showed that the forecasting system yielded a six times return on investment by preventing future displacement and humanitarian needs from arising.