Evaluation of Key Parameters Using Deep Convolutional Neural Networks for Airborne Pollution (PM10) Prediction

Table 5

Final ensemble for MLP architectures.

Architecture	Activation function	Number of neurons	RMSE	MAPE	IOA	R^2
MLP	Linear	10	19.7213	0.239	0.863	0.582
		12
		14
		16
		20
	Softplus	14

The architectures that presented a better combined performance are presented here, next to the metrics of the ensemble. The linear activation function is the most repeated in this ensemble.