Effective Estimation of Hourly Global Solar Radiation Using Machine Learning Algorithms

Table 8

Training and test estimation results of the most successful models that were developed by using the MFFNN algorithm according to input data groups of both provinces.

	Training models								Testing models
	Data	Hidden		Training data MSE	Test data MSE	Validation data MSE	Model MSE	Model	RMSE	MAE	SMAPE (%)
		Layer	Neurons
Isparta	GR1	1	18	0.0198	0.0202	0.0202	0.0201	0.7970	0.1392	0.1109	15.27	0.6547
	GR2		28	0.0029	0.0031	0.0034	0.0030	0.9724	0.0559	0.0372	8.36	0.9444
	GR3		48	0.0024	0.0029	0.0026	0.0025	0.9768	0.0536	0.035	7.77	0.9488
	GR4		32	0.0212	0.0217	0.0216	0.0216	0.7803	0.1446	0.1124	15.29	0.6275
	GR5		21	0.0059	0.0057	0.0063	0.0059	0.9442	0.0754	0.0562	8.97	0.8988
	GR-6		25	0.0027	0.0030	0.0029	0.0028	0.9745	0.0542	0.0363	8.39	0.9477
Kahramanmaras	GR1	1	30	0.0134	0.0137	0.0141	0.0137	0.9049	0.1181	0.0897	14.25	0.8138
	GR2		40	0.0021	0.0026	0.0026	0.002	0.9845	0.0508	0.034	7.79	0.9656
	GR3		27	0.0029	0.0031	0.0028	0.0030	0.9800	0.0556	0.0385	8.23	0.9587
	GR4		25	0.0145	0.0145	0.0150	0.0149	0.8959	0.1234	0.0928	14.37	0.7969
	GR5		44	0.0027	0.0030	0.0033	0.0028	0.9812	0.0558	0.0386	8.29	0.9585
	GR-6		39	0.0028	0.0028	0.0029	0.0029	0.9805	0.0551	0.0383	8.11	0.9595