Table 2.

Assessment of the performance of the models.

Model			Subset 1a, RMSEb					Subset 2a, RMSE					Subset 3a, RMSE					Subset 4a, RMSE
			Training set		Test set		Training set			Test set		Training set			Test set		Training set			Test set
Predictions of new daily COVID-19 cases
	GLM1c	62.22		66.92		53.04			32.70d		48.01			378.94		85.75			219.22
	GLM2e	43.71		29.29d		36.80			569,037.92		48.19			495.88		120.76			429.51
	GLM3f	982.42		587.65		329.49			8,247,155.77		184.59			543.20		330.15			4161.61
	LR1g	58.57		60.17		50.90			44.92		48.20			373.58		85.09			216.22d
	LR2h	56.88		79.57		49.41			78.32		48.00			366.19d		84.52			216.70
	LR3i	56.51		69.13		50.90			44.92		48.20			373.58		84.42			217.81
Predictions of new daily COVID-19 deaths
	GLM1	3.10		4.89		2.52			1.04		2.08			6.79		2.80			4.89
	GLM2	3.24		5.52		2.71			0.47		2.23			7.65		2.82			5.26
	GLM3	3.25		3.79d		2.72			0.19d		2.24			17.02		3.81			4.64d
	LR1	3.05		4.95		2.62			1.71		2.16			5.21		2.75			5.23
	LR2	3.04		4.50		2.61			0.70		2.19			4.82d		2.75			5.38
	LR3	3.05		4.95		2.62			1.71		2.16			5.23		2.75			5.23

^aSubsets 1 to 4: 3, 6, 12, and 18 months after the first case was reported in South Korea, respectively.

^bRMSE: root mean square error.

^cGLM1: generalized linear model with a normal distribution.

^dThe lowest RMSE value in the test subset.

^eGLM2: generalized linear model with a Poisson distribution.

^fGLM3: generalized linear model with a negative binomial distribution.

^gLR1: linear regression model with lasso regularization.

^hLR2: linear regression model with adaptive lasso regularization.

ⁱLR3: linear regression model with elastic net regularization.