Generalizability of machine learning in predicting antimicrobial resistance in E. coli: a multi-country case study in Africa

Table 3 Performance of different machine-learning methods for predicting AMR on Africa data

Antibiotic	Model	Accuracy	Precision	Recall	f1_score	auc_roc
CIP	Logistic Regression	0.55	0.59	0.16	0.25	0.48
	Random Forest	0.50	0.48	0.68	0.56	0.53
	SVM	0.50	0.31	0.05	0.08	0.38
	Gradient Boosting	0.52	0.49	0.32	0.39	0.49
	XGBoost	0.57	0.67	0.17	0.27	0.56
	LightGBM	0.55	0.62	0.12	0.20	0.51
	CatBoost	0.56	0.61	0.17	0.27	0.58
	Feed-Forward NN (Keras)	0.56	0.71	0.12	0.21	0.49
AMP	Logistic Regression	0.94	0.95	0.99	0.97	0.60
	Random Forest	0.38	0.93	0.38	0.54	0.34
	SVM	0.86	0.94	0.91	0.93	0.57
	Gradient Boosting	0.59	0.96	0.59	0.73	0.68
	XGBoost	0.82	0.96	0.84	0.90	0.62
	LightGBM	0.34	0.89	0.35	0.50	0.38
	CatBoost	0.39	0.98	0.36	0.53	0.64
	Feed-Forward NN (Keras)	0.66	0.95	0.67	0.79	0.62
CTX	Logistic Regression	0.39	0.96	0.29	0.45	0.70
	Random Forest	0.20	0.88	0.08	0.14	0.61
	SVM	0.25	1.00	0.12	0.21	0.68
	Gradient Boosting	0.22	1.00	0.09	0.16	0.57
	XGBoost	0.44	0.94	0.36	0.52	0.65
	LightGBM	0.45	0.95	0.38	0.54	0.63
	CatBoost	0.24	0.86	0.13	0.23	0.55
	Feed-Forward NN (Keras)	0.15	0.00	0.00	0.00	0.63

ISSN: 1471-2164