A Multiple Kernel Learning Approach for Air Quality Prediction

Table 6

The optimal parameter settings of the algorithms.

Algorithm	Parameter	Algorithm	Parameter
ARIMA	order (p,d,q): (2,0,2)
RFC	n_estimators: 400; max_depth: 9; max_features: 11; min_samples_split: 95; min_samples_leaf: 71	SVC_linear	C: 400
		SVC_rbf	C: 300,; gamma: 0.02
		SVC_sig	C: 100,; gamma: 0.13,; coef0: 400
MLP	hidden_layer_sizes: (20, 40, 20); activation: ‘relu’; solver: ‘adam’	SVC_poly	C: 100,; degree: 2,; gamma: 0.04,; coef0: 900
MKSVC	Kernel weights of linear, rbf, poly and sig kernels: (0.999, 0.212, 0,134, 0.00009)