A machine learning-based exploration of resilience and food security

Abstract

Leveraging advancements in remote data collection and using the Food Insecurity Experience Scale (FIES) as a proxy measure of resilience, we show that machine learning models (such as Gradient Boosting Classifier, eXtreme Gradient Boosting, and Artificial Neural Networks), can predict resilience with relatively high accuracy (up to 81%). Key household-level predictors include access to financial institutions, asset ownership, the adoption of agricultural mechanization as evidenced by the use of tractors, the number of crops cultivated, and ownership of nonfarm enterprises. Our analysis offers insights to researchers and policymakers interested in the development of targeted interventions to bolster household resilience.