Table 2.

Evaluation metrics of the 5 machine learning models for the current week forecasts (nowcast), 1-week forecasts, 2-week forecasts, and 3-week forecasts in 2018.

			Outpatient influenza-like illness visits					Emergency influenza-like illness visits
	Model	RMSEa		MAPEb, %	Hit rate	Pearson correlation coefficient	RMSE		MAPE, %	Hit rate	Pearson correlation coefficient
Nowcast (current week)
	ARIMAc	6422.8		5.5	0.727	0.958	1125.3		5.7	0.782	0.965
	RFd	6472.1		5.4	0.709	0.957	1305.8		6.2	0.709	0.952
	SVRe	5343.2		5.8	0.655	0.969	2079.6		8.0	0.582	0.875
	XGBf	7384.6		5.8	0.691	0.943	1643.6		6.2	0.673	0.923
	Ensemble	6170.7		5.3	0.600	0.962	1751.0		6.5	0.727	0.912
1-week
	ARIMA	9874.7		9.0	0.741	0.897	1707.6		7.8	0.704	0.919
	RF	8644.1		7.2	0.833	0.921	1861.2		7.9	0.759	0.899
	SVR	7330.1		7.9	0.778	0.942	2643.7		10.6	0.759	0.798
	XGB	9738.7		9.0	0.741	0.903	2353.3		8.3	0.685	0.836
	Ensemble	9156.8		7.5	0.796	0.911	2363.0		8.4	0.722	0.832
2-week
	ARIMA	11630.6		11.8	0.811	0.851	1922.2		8.6	0.755	0.893
	RF	10082.3		9.4	0.811	0.893	1975.5		7.7	0.736	0.888
	SVR	9292.7		10.1	0.830	0.905	3031.0		12.8	0.566	0.730
	XGB	11262.8		11.3	0.755	0.873	2371.6		7.9	0.698	0.843
	Ensemble	10078.8		8.6	0.774	0.889	2389.4		8.6	0.755	0.835
3-week
	ARIMA	13656.7		13.5	0.865	0.787	1875.4		9.5	0.788	0.898
	RF	10258.0		10.1	0.769	0.892	2041.9		8.8	0.692	0.877
	SVR	9439.7		10.9	0.885	0.904	3106.0		13.5	0.596	0.721
	XGB	12789.3		13.0	0.788	0.830	2259.6		7.6	0.692	0.890
	Ensemble	9160.9		8.8	0.904	0.908	2478.4		9.6	0.769	0.814

^aRMSE: root mean squared error.

^bMAPE: mean absolute percentage error.

^cARIMA: autoregressive integrated moving average.

^dRF: random forest.

^eSVR: support vector regression.

^fXGB: extreme gradient boosting.