. 2022 Jul 20;10(7):e34504. doi: 10.2196/34504

Table 5.

DM test results of testing data set for different data sets.

Test model (with internet data)	Reference model (without internet data)^a
Test model (with internet data)	GLM^b	ARIMAX^c	ANN^d	SVM^e	RF^f	LSTM^g	ELM^h
GLM	2.4848 (<.001)	0.4041 (.31)	0.2797 (.37)	0.2314 (.40)	1.8806 (.22)	2.1002 (.31)	0.8635 (.16)
ARIMAX	1.5701 (.04)	2.5818 (<.001)	0.4968 (.28)	1.7698 (.12)	0.7756 (.20)	1.5748 (.24)	0.0337 (.51)
ANN	2.2546 (<.001)	1.7547 (.02)	4.1945 (<.001)	2.8276 (<.001)	3.4393 (<.001)	1.2432 (.09)	4.4291 (<.001)
SVM	2.244 (<.001)	1.7374 (.02)	6.0597 (<.001)	2.3394 (<.001)	1.6767 (.02)	0.8602 (.07)	3.2791 (<.001)
RF	1.7886 (.02)	1.9591 (.02)	2.7599 (<.001)	1.8785 (.02)	2.3097 (<.001)	0.5800 (.05)	2.3075 (<.001)
LSTM	0.8556 (.01)	3.3685 (<.001)	1.3620 (.07)	1.6441 (.04)	1.0560 (.12)	2.4263 (<.001)	1.9995 (.02)
ELM	2.2546 (<.001)	1.7547 (.03)	4.1946 (<.001)	2.8276 (<.001)	3.4394 (<.001)	2.175 (.01)	4.4291 (<.001)

^aValues are presented as the Diebold-Mariano statistic. The P value modified by false discovery rate is in brackets. The significance level is .05.

^bGLM: generalized linear model.

^cARIMAX: ARIMA with explanatory variables.

^dANN: artificial neural network.

^eSVM: support vector machine.

^fRF: random forest.

^gLSTM: long short-term memory.

^hELM: extreme learning machine.