Stunting in children continues to persist in many developing countries, particularly in sub-Saharan Africa. A multiple-country comparative analysis provides an opportunity for learning what works as far as tackling child stunting in the sub-Saharan Africa region. The objective is to identify the most effective model and key predictors for stunting prediction using machine learning (ML) techniques in four East African countries (Kenya, Tanzania, Malawi and Zambia). Based on the best predictive model, XGBoost, the most important predictors of stunting status differed in the four countries but included child, maternal and household factors.