Table 4.

DM^a test results of testing data set for same data set.

Test model			Reference modelb,c
			GLMd		ARIMAXe		ANNf		SVMg		RFh		LSTMi
Patient arrivals + traditional variables
	ELM	2.8297 (<.001)		3.0624 (<.001)		2.012 (<.001)		1.8178 (<.001)		2.8481 (<.001)		2.1002 (<.001)
	GLM			0.2935 (.31)		0.86595 (.11)		1.0707 (.06)		0.7643 (.09)		0.1663 (.54)
	ARIMAX					0.64435 (.17)		1.0691 (.06)		0.3957 (.23)		0.4876 (.68)
	ANN							0.13244 (.38)		0.2746 (.45)		1.9512 (.01)
	SVM									0.5823 (.27)		0.8714 (.12)
	RF											1.0045 (.08)
Patient arrivals + traditional variables + internet search index
	ELM	2.5062 (<.001)		3.79 (<.001)		2.0047 (<.001)		2.0325 (<.001)		2.0476 (<.001)		1.6659 (.02)
	GLM			0.32675 (.30)		1.1462 (.12)		1.6064 (.07)		1.0467 (.09)		0.3647 (.64)
	ARIMAX					1.7314 (.06)		2.2885 (.05)		1.5946 (.11)		1.2671 (.07)
	ANN							0.14419 (.40)		0.2104 (.49)		1.2304 (.08)
	SVM									0.2593 (.36)		1.4391 (.04)
	RF											1.2992 (.06)

^aDM: Diebold-Mariano.

^bThe P value modified by false discovery rate is given in brackets. The significance level is .05.

^cValues are presented as the Diebold-Mariano statistic (P value modified by false discovery rate).

^dGLM: generalized linear model.

^eARIMAX: ARIMA with explanatory variables.

^fANN: artificial neural network.

^gSVM: support vector machine.

^hRF: random forest.

ⁱLSTM: long short-term memory.