In 2011, UNHCR refugee camps in Ethiopia were scrambling. Neighboring Somalia was experiencing a severe drought, which, in combination with lasting conflict, left Somalia’s people impacted by the worst famine the country had seen in 25 years (UNHCR, 2019). Untenable living conditions drove hundreds of thousands of Somalis to seek refuge in neighboring countries, desperate to save themselves and their families.
Unprepared for the influx of people, UNHCR refugee camps were left struggling to meet the needs of existing camp residents as well as the nearly 2,000 daily arrivals. Humanitarian workers, perpetually stretched thin during regular times, were rushing to process new arrivals, provide food to starving children and parents, and ensure everyone had a place to sleep. The Dollo Ado camp team recalls spending their entire yearly budget for nutritional screening and protection support within a few short months (UNHCR, 2019). Eventually, the crisis abated, but in 2017 yet another drought seized the country. Haunted by their experience in 2011, teams on the ground requested UNHCR’s assistance in developing a method to predict migration flows so refugee camp workers could ensure adequate staff and resources to meet the needs of every individual that came through their doors.
Understanding the urgency of the situation, UNHCR began developing Project Jetson, a machine learning program that utilizes supervised machine learning to predict Somali refugee and IDP migration patterns up to one month in advance.