Project: Transfer learning for socioeconomic data (SocioSAT)
Project: Transfer learning for socioeconomic data (SocioSAT)
Farchana refugee camp, eastern Chad.
UNHCR is working with partners to develop SocioSAT, an earth observation and machine learning system that enables high-frequency assessment of the socioeconomic living conditions of people affected by forced displacement, to help countries plan effective responses and policies.
About the SocioSAT project
SocioSAT is an innovative project that combines earth observation and machine learning to estimate socioeconomic conditions in settlements for:
- refugees
- internally displaced people
- host communities.
It is a joint data project led by UNHCR in collaboration with Linköping University's AI & Global Development Lab and funded by the World Bank-UNHCR Joint Data Center on Forced Displacement.
SocioSAT will be the first system of its kind to apply transfer learning from models trained on multi-year, multi-country general population surveys to forced displacement contexts. Household surveys provide an important data foundation for sustainable responses to forced displacement. However, high-quality representative surveys can only be implemented infrequently in many contexts.
This project aims to address the timeliness of this data, through a complementary estimation pipeline to fill gaps between household survey data collection rounds. The project adapts a model trained on Demographic and Health Survey (DHS) data from approximately 1.2 million households across 36 African countries to the unique settings of displacement. It combines daytime satellite imagery, nighttime light data and UNHCR surveys such as the Forced Displacement Survey (FDS) and Results Monitoring Survey (RMS) to produce settlement-level welfare estimates.
A refugee home in Gorom Refugee Camp, Central Equatoria State, South Sudan.
Text and media 5
The system will support UNHCR operations, partners and host governments with timely, granular socioeconomic information to:
- better prioritize resources
- monitor living conditions
- respond more effectively to evolving needs on the ground.
The project will support UNHCR's digital transformation strategy and AI Approach, which seek to optimize the delivery of protection programmes through innovative data-driven approaches.
It is being piloted in selected countries in the Sub-Saharan Africa region, with planned extensions to other regions.
Contact us
For more information about the project, please contact UNHCR’s Statistics, Data Science and Survey Section at [email protected].