Classification tree and random forest model to predict under-five malnutrition in Bangladesh

Abstract

Malnutrition is one of the leading causes of morbidity and mortality in children under the age of five in most developing countries like Bangladesh. The main objective of this study is to design a model that predicts the nutritional status of under-five children using tree based model and classical approach. This study used secondary data from Bangladesh Demographic and Health Survey 2014 for 7,886 children. Decision tree based model like classification tree, random forest and classical model like multiple binary logistic regression model are fitted to assess the association of malnutrition of children with potential socioeconomic and demographic factors. In this study, predictive model is developed using random forest having an accuracy of 70.1% & 72.4% and area under receiver operating characteristic curve of 69.8% and 70% for stunting and underweight respectively. The prevalence of stunting and underweight are found 36.5% and 33% respectively among under-five children and higher in rural setting than in urban areas. Similarly, wealth index, exposure of mother to the mass media, age of child, size of child at birth, and parents’ education are significantly associated with stunting and underweight of children.