Malnutrition detection in infants using machine learning approach

Abstract

Balanced nutrition is the main source of energy. It is necessary for healthy life of people. Healthy nutrients enable cells to perform their regular activities at pace. Deficiency of proper nutrition while birth causes various complications in further life. These complications include wasting, stunting, edema, mental illness, low immune system, ridged or spoon-shaped nails, brittle, dry hair, and underweight etc. Malnutrition is a condition that occurs when a person consumes a diet that is deficient in one or more major nutrients, or has too many of them. Marasmus, kwashiorkor and intermediate states of marasmus-kwashiorkor are included in the term Protein-Energy Malnutrition (PEM) disorders. PEM is the cause of underweight (low weight for age), stunting (low height for age), and wasting (low weight for height). In India, stunting affects 48% of infants under five years age, wasting affects 20%, and underweight affects 43%. Most children suffering from undernutrition in mild to moderate forms are unnoticed in India, which affects their growth at early ages. Detecting malnutrition at early stage reduces further healthcare cost and improve health outcome. To alleviate the problem of malnutrition, this paper describes a decision tree model for classification of infants being between the ages of 0 and 59 months as normal, acute malnourished or severely malnourished for three categories: Stunting, Wasting and Underweight. In decision tree model, Gini index is adopted as an impurity measure. The accuracy obtained using decision tree for stunting is 82.22%, for wasting 72.23 % and underweight 78.35% using Gini index.