. 2021 May 14;21(10):3414. doi: 10.3390/s21103414

Table 6.

Results for the optimization of the KNN and RF models, for different values of K, Manhattan distance and number of trees (T). Values in bold represent the best results for each model. Grey rows represent the selected parameters.

		Arousal			Valence
Model	Par.	PCC	MAE	RMSE	PCC	MAE	RMSE
KNN $α, β, γ (A l l - S E) +$ $α_{D A} (A l l - H 3)$	K = 1	0.794	0.062	0.163	0.795	0.066	0.172
	K = 3	0.725	0.120	0.175	0.725	0.128	0.185
	K = 5	0.684	0.137	0.185	0.689	0.146	0.194
	K = 7	0.655	0.147	0.192	0.663	0.156	0.201
	K = 11	0.622	0.156	0.199	0.633	0.166	0.208
	K = 21	0.579	0.166	0.208	0.595	0.176	0.217
RF $β, γ (H 1 + W P) +$ $α_{D A} (H 1 + W P + W E)$	T = 50	0.740	0.137	0.176	0.839	0.119	0.156
	T = 100	0.748	0.136	0.175	0.845	0.119	0.155
	T = 500	0.755	0.135	0.174	0.852	0.118	0.153
	T = 750	0.755	0.135	0.174	0.852	0.118	0.153
	T = 1000	0.756	0.135	0.174	0.853	0.118	0.153