PREDICTING RISK OF INADEQUATE MICRONUTRIENT INTAKEWITH TRANSFERABLE MACHINE LEARNING MODELS

Abstract

Identifying populations at risk of inadequate micronutrient intake is useful for governments and humanitarian organizations in low- and middle-income countries to make informed and timely decisions on nutrition relevant policies and programmes. We propose a machine-learning methodological approach using secondary data on household dietary diversity, socioeconomic status, and climate indicators to predict the risk of inadequate micronutrient intake in Ethiopia and in Nigeria. We identify key predictive features common to both countries, and we demonstrate the models transferability from one country to another to predict risk of inadequate micronutrient intake in contexts where nationally representative primary data are unavailable.