<![CDATA[Nutritional status of toddlers is a critical indicator of early childhood growth and health; however, manual identification methods remain prone to inconsistency and delayed decision-making, particularly when applied to large-scale data. The absence of a Hybrid system integrating Machine Learning and Rule-based approaches for an Early Warning System (EWS) of toddler nutritional status represents a research gap that needs to be addressed. This study implements an Artificial Neural Network (ANN) using the Multilayer Perceptron (MLPClassifier) algorithm on a synthetic dataset generated from the WHO child growth Z-score formula, with class distributions of Normal (54.57%), Tall (17.69%), Severely Stunted (16.54%), and Stunted (11.20%). Stepwise experiments were conducted on hidden layer configurations, activation functions, and solvers to determine the optimal model. The best configuration was achieved using hidden layers (64,32), ReLU activation, and LBFGS solver, yielding an accuracy of 0.9954 and a Macro F1-Score of 0.9935. Validation through 5-fold cross-validation produced a mean accuracy of 0.9955 with a standard deviation of 0.0005, confirming model stability and absence of overfitting. The model was integrated into a Rule-based EWS to provide early risk-level classification of toddler growth status. This Hybrid ANN and Rule-based combination proves effective as a decision support system in child healthcare.]]>
Copyrights © 2026
