Skip to main content

View full-text article in PMC

. Author manuscript; available in PMC: 2021 Dec 1.

Published in final edited form as: Environ Int. 2020 Sep 24;145:106143. doi: 10.1016/j.envint.2020.106143

Table 2.

Performance statistics of 100 trained models in ensemble learning

	Training		Validation^a		Testing^b
	Mean	Range	Mean	Range	Mean	Range

Sample size	17,629		4,407		12,776
R² (range)	0.94	(0.77, 0.97)	0.82	(0.75, 0.85)	0.82	(0.70, 0.85)
RMSE (μg/m³)	1.54	(1.03, 3.02)	2.70	(2.49, 3.24)	2.70	(2.46, 3.49)

^a

: the samples not used to train the models but used to validate the model (adjusting the hyper-parameters to get an optimal effect);

^b

: the samples used to test the trained models (not used in training and validation).