Table 7.

Prediction performance of four different approaches for mental illnesses.

Illness	Variable Selection	Under Sampling	Model	Accuracy	F1-Score	Sensitivity	Specificity	AUC
Depression	A-$R^2$	Yes	LGBM	0.7801	0.7913	0.8338	0.7265	0.8626
	No	Yes	LGBM	0.7648	0.7745	0.8083	0.7214	0.8518
	No	No	LGBM	0.8530	0.9199	0.9903	0.0615	0.7648
	LR	No	LGBM	0.8524	0.9201	0.9974	0.0160	0.7567
Anxiety	A-$R^2$	Yes	LGBM	0.8293	0.8286	0.8251	0.8335	0.8845
	No	Yes	LGBM	0.8242	0.8231	0.8178	0.8306	0.8775
	No	No	LGBM	0.8558	0.9100	0.9380	0.5701	0.8318
	LR	No	LGBM	0.8550	0.9091	0.9331	0.5833	0.8289
Schizophrenia	LR	Yes	RF	0.8759	0.8770	0.8976	0.8514	0.9292
	No	Yes	RF	0.8779	0.8799	0.8983	0.8544	0.9268
	No	No	XGB	0.9962	0.9981	1.0000	0.0000	0.7423
	LR	No	XGB	0.9962	0.9981	1.0000	0.0000	0.7361
Disease	No	Yes	LGBM	0.8663	0.8772	0.9550	0.7776	0.9159
	A-$R^2$	Yes	XGB	0.8670	0.8788	0.9646	0.7695	0.9130
	No	No	LGBM	0.8565	0.9035	0.9971	0.5659	0.8522
	LR	No	LGBM	0.8555	0.9028	0.9957	0.5656	0.8515