Gridding forced displacement using semi-supervised learning

This first CLIFDEW technical brief presents an innovative methodology for mapping forced displacement throughout Sub-Saharan Africa. The study uses semi-supervised learning to break down refugee statistics into 0.5-degree grid cells by combining UNHCR's ProGres registration data with satellite-derived building footprints and settlement information. The outcome is a high-resolution dataset encompassing over 10 million records, which reveals localized displacement patterns with over 92% accuracy, providing new insights into the factors driving displacement and facilitating more targeted humanitarian responses.