Using Machine Learning for Programmatic Synthesis
POSTED December 29, 2021|
Agriculture and Food Security
Choose a file to download:
By Brooke Jardine – NORC at the University of Chicago for USAID Research Technical Assistance Center (RTAC)
In the summer of 2020, the Analysis and Learning Division (ALD) at the USAID Bureau for Resilience and Food Security partnered with RTAC to test machine learning (ML) methods for aggregating and synthesizing information contained in water sector documents to inform approaches by the Center for Water (C4W) and achieve development objectives. These activities were seeking to answer two questions: 1) To what extent are USAID water, sanitation, and hygiene (WASH) field activities aligned with the strategic and programmatic approaches defined by C4W as best practices? and 2) What other approaches are most used in field activities?
Through a process of data collection, cleaning, analysis, modeling, and deployment the research team aimed at providing an overview of lessons learned and recommendations to those at USAID that are interested in using ML to analyze programming information. Recommendations from this process include understanding how appropriate ML is for responding to certain questions; successful planning prior to implementing a ML plan; and clearly defining what constitutes success at different stages of the ML process; among others.
Available resources in this page include a research report and a learning brief for a non-specialist audience.