Table 2.

Average performance of the trained models classifying malnutrition using all (n = 134) features.

	Decision tree	KNN	SVM	Logistic regression	Naive Bayes	Random Forest	AdaBoost	LGBM	XGBoost
Sensitivity	0.875 ± 0.107	0.491 ± 0.152	0.807 ± 0.123	0.903 ± 0.096	0.696 ± 0.140	0.849 ± 0.107	0.863 ± 0.109	0.888 ± 0.101	0.849 ± 0.112
Specificity	0.909 ± 0.070	0.954 ± 0.046	0.852 ± 0.077	0.843 ± 0.078	0.841 ± 0.083	0.907 ± 0.065	0.941 ± 0.055	0.935 ± 0.054	0.935 ± 0.053
F1-score	0.85 ± 0.085	0.608 ± 0.139	0.766 ± 0.09	0.815 ± 0.076	0.688 ± 0.110	0.833 ± 0.082	0.870 ± 0.079	0.879 ± 0.075	0.856 ± 0.078

Performance was evaluated for 100 × 10-fold cross-validation iterations (n = 1,000). The highest value for each performance metric across all algorithms is highlighted. Data are shown as mean ± standard deviation. KNN, K-nearest neighbors; SVM, support vector machine; AdaBoost, adaptive boosting; LGBM, light gradient boosting machine; XGBoost, eXtreme gradient boosting. Bold values indicate statistical significance (p < 0.05).