Table 2 The effect of different machine learning algorithms on model prediction performance using SMOTE sampling method.

Inputing method	Sampling method	Selection methods	Model name	AUC	Accuracy	Precision	Recall	F1 Score	Specificity
RF	SMOTE	BOR	XGB	0.5286	0.5092	0.2751	0.4765	0.3488	0.5217
RF	SMOTE	BOR	RF	0.5296	0.5092	0.2751	0.4765	0.3488	0.5217
RF	SMOTE	BOR	SVC	0.5351	0.5074	0.2741	0.4765	0.3480	0.5191
RF	SMOTE	BOR	KNN	0.4893	0.7203	0.3333	0.0134	0.0258	0.9897
RF	SMOTE	BOR	CB	0.5301	0.5092	0.2751	0.4765	0.3488	0.5217
RF	SMOTE	BOR	LR	0.5149	0.4722	0.2763	0.5637	0.3708	0.4373
RF	SMOTE	BOR	MLP	0.5995	0.2759	0.2759	1	0.4325	0
RF	SMOTE	BOR	SGD	0.5153	0.4351	0.3020	0.7986	0.4383	0.2966
RF	SMOTE	BOR	DT	0.5242	0.5037	0.2551	0.4161	0.3163	0.5370
RF	SMOTE	LA	XGB	0.9055	0.8444	0.7685	0.6241	0.6888	0.9283
RF	SMOTE	LA	RF	0.9010	0.8129	0.7448	0.4899	0.5910	0.9360
RF	SMOTE	LA	SVC	0.8926	0.8314	0.7042	0.6711	0.6872	0.8925
RF	SMOTE	LA	KNN	0.7402	0.7074	0.4771	0.630872	0.5433	0.7365
RF	SMOTE	LA	CB	0.9124	0.8425	0.7461	0.651007	0.6953	0.9156
RF	SMOTE	LA	LR	0.8811	0.8074	0.6203	0.778523	0.6904	0.8184
RF	SMOTE	LA	MLP	0.8439	0.8092	0.6796	0.583893	0.6281	0.8951
RF	SMOTE	LA	SGD	0.9013	0.8259	0.6519	0.791946	0.7151	0.8388
RF	SMOTE	LA	DT	0.8556	0.7925	0.6069	0.704698	0.6521	0.8260
RF	SMOTE	RFE	XGB	0.9097	0.8537	0.7243	0.758389	0.7409	0.8900
RF	SMOTE	RFE	RF	0.9097	0.8592	0.7417	0.751678	0.7466	0.9002
RF	SMOTE	RFE	SVC	0.8967	0.8425	0.6860	0.791946	0.7352	0.8618
RF	SMOTE	RFE	KNN	0.8758	0.8277	0.7692	0.536913	0.6324	0.9386
RF	SMOTE	RFE	CB	0.9170	0.8388	0.6684	0.825503	0.7387	0.8439
RF	SMOTE	RFE	LR	0.8807	0.8055	0.6182	0.771812	0.6865	0.8184
RF	SMOTE	RFE	MLP	0.9068	0.8407	0.6800	0.798658	0.7345	0.8567
RF	SMOTE	RFE	SGD	0.9012	0.8240	0.6377	0.838926	0.7246	0.8184
RF	SMOTE	RFE	DT	0.8812	0.8629	0.7516	0.751678	0.7516	0.9053

SMOTE synthetic minority oversampling technique, BOR boruta screening, RFE recursive feature elimination, LA lasso screening, XGB extreme gradient boosting, SVC support vector classify, KNN K-nearest neighbor, CB category boosting, LR logistic regression, MLP multilayer perceptron, SGD stochastic gradient descent, DT decision tree, AUC area under the curve.

Search