PREDICTING UNDERNUTRITION RISK FACTORS USING MACHINE LEARNING TECHNIQUES IN NIGERIAN UNDER FIVE CHILDREN

Childhood undernutrition remains a critical global health challenge with far-reaching consequences. This study developed and compare predictive models for three key indicators of under-five child undernutrition in Nigeria: stunting, wasting, and underweight. By leveraging various machine learning (ML) algorithms, we identified the most significant socio-demographic determinants of these nutritional outcomes. The study utilizes data from the Nigerian Multiple Indicator Cluster Survey (MICS6) 2021, a comprehensive nationwide survey. Four ML algorithms were employed to predict the risk factors for the under-5 child nutritional status: k-nearest neighbors (KNN), random forest (RF), decision tree (DT), and logistic regression (LR). These models were evaluated and compared based on their predictive performance. The study encompasses households across Nigeria, providing a broad representation of the country's diverse population. Preliminary analysis reveals significant regional disparities in the prevalence of stunting, wasting, and underweight among Nigerian children. Among the four ML algorithms tested, the KNN model demonstrated superior predictive capability across all evaluation metrics, including accuracy (89.65%, 95.00%, 80.04%), precision (91.02%, 87.79%, 93.53), recall (99.98%, 100%, 100%), and F1-score (94.44%, 98.83%, 88.60%). This outperformance was consistent for all three undernutrition indicators. The KNN model, identified as the most effective predictor, highlighted several key determinants of childhood undernutrition. These factors varied somewhat across the three outcomes but commonly included: household wealth index, geopolitical zone, source of drinking water, child age, birth size, mother's education level, and residential area (urban/rural) among others. This study demonstrates the efficacy of machine learning approaches, particularly the KNN algorithm, in predicting and understanding the determinants of children undernutrition in Nigeria. The findings provide valuable insights for targeted interventions and policy formulation. These results can inform evidence-based decision-making and resource allocation in Nigeria's efforts to improve child nutrition and achieve related Sustainable Development Goals.