<![CDATA[Stunting is a chronic nutritional problem that seriously impacts child growth and development. This study aims to compare the performance of the Naïve Bayes and K-Nearest Neighbors (KNN) algorithms in predicting stunting in toddlers in Banjaran District. The dataset consists of 12,000 toddler data points with three main features: age, gender, and height. The research employed a quantitative approach by applying machine learning algorithms. The SMOTE oversampling technique was applied only to the training data to avoid data leakage, and 5-fold cross-validation was used. A K-value of 3 was selected for the final KNN model based on validation curve analysis to prevent overfitting. The results show that KNN significantly outperformed Naïve Bayes across all evaluation metrics. The Naïve Bayes model yielded an accuracy of 67.50%, precision of 50.87%, recall of 61.38%, F1-score of 55.63%, specificity of 70.54%, and an AUC score of 75.71%. Meanwhile, the KNN (K=3) model achieved an accuracy of 99.11%, precision of 98.08%, recall of 99.25%, F1-score of 98.66%, specificity of 99.03%, and an AUC score of 99.65%. The performance difference between the two models was confirmed by McNemar's Test with a p-value < 0.05, indicating a statistically significant difference. The low performance of Naïve Bayes was attributed to the violation of the feature independence assumption, particularly the high correlation between age and height (r ≈ 0.87). In conclusion, KNN is the more appropriate algorithm for stunting prediction on this dataset. However, the limitation of features suggests the need for further research with additional variables and external validation before wider-scale implementation.]]>
