Table 2.

Model performance in 5-fold cross-validation

	Accuracy (95% CI)	Weighted F1 score (95% CI)	Micro-averaged F1 (95% CI)
L1	0.568 (0.559–0.578)	0.551 (0.537–0.565)	0.568 (0.559–0.578)
L2	0.606 (0.591–0.621)	0.556 (0.536–0.577)	0.606 (0.591–0.621)
L1-L2	0.607 (0.593–0.621)	0.562 (0.543–0.582)	0.607 (0.593–0.621)
RF (multiclass)	0.638 (0.622–0.653)*	0.612 (0.594–0.630)*	0.638 (0.623–0.654)*
RF (1-versus-rest)	0.649 (0.632–0.666)*	0.621 (0.601–0.640)*	0.649 (0.632–0.666)*

Abbreviations: CI, Confidence Interval. L1, L1 penalized logistic regression; L2, L2 penalized logistic regression; L1-L2, L1-L2 penalized logistic regression; RF, random forest.

^*P-values <0.05 within each of the 3 performance measures for comparisons between RF (multiclass) and, L1, L2, and L1-L2 LR models and for comparisons between RF (1-versus-rest) and, L1, L2, and L1-L2 LR models.