Title : Leveraging convolutional neural networks for multi-dimensional classification of nutritional status using anthropometric indicators in child health
Abstract:
Child malnutrition remains a critical global health issue, requiring advanced and efficient methods for accurate classification and early intervention. This study explores the application of Convolutional Neural Networks (CNNs) for multi-dimensional classification of nutritional status using anthropometric indicators in child health. A dataset containing key demographic and biometric variables, including sex, age, weight, height, BMI, household size, maternal characteristics, and environmental factors, was preprocessed and transformed for deep learning analysis.
The CNN model was trained on a structured dataset, where BMI categories served as classification labels—underweight, normal weight, overweight, and obese. Data preprocessing involved label encoding categorical variables, normalizing numerical features, and reshaping input dimensions to fit a one-dimensional CNN framework. The model architecture consisted of four convolutional layers, two max pooling layers, and three fully connected layers, optimized with the Adam optimizer and categorical cross-entropy loss function.
After training for 50 epochs with a batch size of 32, the model achieved a training accuracy of 96.85% and a validation accuracy of 94.21%, with a final loss of 0.113 on the test dataset. The precision, recall, and F1-score across all classes averaged 93.75%, 94.12%, and 93.92%, respectively, highlighting the model’s robustness in identifying nutritional status from anthropometric indicators. These findings shows the potential of deep learning in automating child malnutrition assessments, enabling healthcare professionals and policymakers to leverage AI-driven tools for improved nutritional surveillance and intervention planning.
Keywords: Convolutional Neural Networks, Nutritional Classification, Anthropometric Indicators, Child Health, Deep Learning
