Effective Estimation of Hourly Global Solar Radiation Using Machine Learning Algorithms

Table 10

Estimation results of the two most successful models that were developed by using a KNN algorithm according to input data groups of both provinces.

Target provinces		Kahramanmaras				Isparta
Data groups	The number of neighbors ()	RMSE	MAE	SMAPE (%)		RMSE	MAE	SMAPE (%)
GR1	6	0.1221	0.0903	14.25	0.8022	0.1447	0.1140	15.53	0.6274
	10	0.1205	0.0896	14.20	0.8069	0.1434	0.1131	15.42	0.6338
GR2	4	0.0618	0.0427	8.91	0.9490	0.0692	0.0481	9.34	0.9151
	6	0.0605	0.0419	8.84	0.9511	0.0693	0.0485	9.38	0.9150
GR3	4	0.0631	0.0437	9.08	0.9470	0.0646	0.0438	8.88	0.9261
	6	0.0631	0.0439	9.13	0.9469	0.0648	0.0450	9.17	0.9258
GR4	6	0.1288	0.0956	14.44	0.7798	0.1506	0.1157	15.42	0.5980
	10	0.1257	0.0937	14.30	0.7898	0.1463	0.1134	15.26	0.6194
GR5	4	0.0645	0.0446	9.37	0.9445	—	—	—	—
	6	0.0649	0.0448	9.32	0.9439	0.0820	0.0614	9.33	0.8805
	10	—	—	—	—	0.0811	0.0610	9.32	0.8829
GR6	3	0.0718	0.0487	10.34	0.9318	—	—	—	—
	4	0.0708	0.0482	10.32	0.9335	0.0728	0.0514	9.79	0.9065
	6	—	—	—	—	0.0730	0.0522	10.07	0.9062