Table 2.

Comparing the best machine learning classifier and DA considering 11 drugs

Drugs	DA			Best method
	Sensitivity	Specificity	AUC	Feature set + Classifier	Sensitivity	Specificity	AUC
INH	91.95 ± 1.04	98.71 ± 0.22	94.95 ± 0.54	F1 + LR – L2	92.19◊ ± 0.94	98.38 ± 0.29	97.89◊ ± 0.38
EMB	83.31 ± 1.62	95.17 ± 0.38	89.24 ± 0.85	F1 + LR – L2	92.12◊ ± 0.98	91.89 ± 0.84	96.25◊ ± 0.54
RIF	91.70 ± 1.19	98.73 ± 0.22	95.22 ± 0.59	F1 + LR – L2	92.27◊ ± 1.25	97.45 ± 0.63	98.08◊ ± 0.32
PZA	43.11 ± 2.97	98.46 ± 0.27	70.78 ± 1.46	F1 + LR – L2	88.12◊ ± 2.65	88.91 ± 1.66	93.89◊ ± 0.80
SM	82.80 ± 1.90	97.19 ± 0.44	89.99 ± 0.99	F1 + LR – L2	87.40◊ ± 1.98	94.15 ± 1.23	95.15◊ ± 0.56
AK	65.21 ± 5.32	99.70 ± 0.24	82.46 ± 2.70	F1 + SPCA + LR – L2	77.23◊ ± 6.96	89.84 ± 3.05	91.37◊ ± 2.36
MOX	62.97 ± 6.60	98.80 ± 0.68	80.89 ± 3.32	F1 + GBT	76.84◊ ± 9.29	87.19 ± 8.21	90.27◊ ± 2.96
OFX	65.07 ± 3.92	99.31 ± 0.28	82.19 ± 1.98	F1 + GBT	79.06◊ ± 6.94	90.88 ± 6.38	92.33◊ ± 1.49
KAN	72.31 ± 5.40	97.61 ± 0.65	84.96 ± 2.68	F1 + LR – L2	80.41◊ ± 6.48	93.48 ± 4.93	92.49◊ ± 2.93
CAP	59.68 ± 5.84	93.87 ± 0.88	76.78 ± 2.96	F1 + SPCA + LR – L2	64.44◊ ± 6.02	92.74 ± 2.52	85.46◊ ± 2.02
CIP	46.65 ± 10.10	99.24 ± 0.89	72.95 ± 5.17	F1 + LR – L2	79.86◊ ± 9.98	85.37 ± 7.65	89.53◊ ± 4.06

Note: Sensitivity, specificity and AUC (mean ± standard error) is reported. Wilcoxon signed-rank test was used to calculate the P-value of each method compared with the DA and ^◊ indicate s P < 0.01.