MACHINE LEARNING APPROACHES TO CHILD STUNTING PREDICTION

Abstract

Children's growth and development are crucial for their well-being and future success, making nutrition a critical aspect of their early years. Malnutrition poses significant risks to children's health, particularly undernutrition, which remains a global concern. Adequate nutrition knowledge is essential for parents to ensure proper food intake for their children, thereby promoting brain development and memory retention. This study aims to address the high prevalence of stunting among children under five in East Aceh, Indonesia, a national priority for evaluating child growth. We employ the random forest method, a supervised machine learning model, to predict stunting in East Aceh and explore its potential as a valuable tool for assessing children's nutritional status. Aceh ranks third in Indonesia for the number of children under five suffering from stunting, a condition closely linked to malnutrition and inadequate nutrition knowledge among parents. Malnourished children are more susceptible to various illnesses compared to their healthier counterparts. Past studies have investigated under-five malnutrition and associated risk factors, primarily relying on classical regression models. However, these traditional models present challenges in handling multicollinearity and a large number of covariates, limiting their accuracy. In contrast, machine learning (ML) methods offer numerous advantages, such as using a larger number of predictors, requiring fewer assumptions, and accommodating multi-dimensional correlations. ML models provide more flexibility in establishing relationships between predictor and outcome variables, making them superior for handling classification problems and prediction tasks. Our study's findings indicate an improvement in the nutritional status of under-five children in East Aceh over the last decades. The prevalence of underweight, stunting, and wasting has decreased significantly. We calculated the Composite Index for Anthropometric Failure (CIAF), which aggregates different forms of anthropometric failure to assess children's nutritional status comprehensively. ML models, particularly the random forest, prove effective in predicting stunting in East Aceh, potentially enabling early intervention to prevent malnutrition. In conclusion, this study highlights the importance of nutrition knowledge among parents in ensuring children's healthy growth. Machine learning techniques offer promising opportunities to predict stunting and undernutrition, aiding public health efforts to combat malnutrition in children.