Analysis of Artificial Neural Network: Architecture, Types, and Forecasting Applications

Table 4

The proposed five input-based multilayer perceptron neural network statistical performance analyses with various hidden neurons for solar irradiance forecasting.

Number of hidden neurons	Error qualifier
	R	MAPE	MSE	MAE	RMSE	MRE	Time (min)
1	1	0.0029	7.1344e-05	0.0063	0.0084	2.9047e-05	3.17
2	1	0.0014	1.8264e-05	0.0031	0.0043	1.4451e-05	4.13
3	1	8.2990e-04	7.3104e-06	0.0018	0.0027	8.2990e-06	2.03
4	1	0.0020	4.3011e-05	0.0043	0.0066	1.9727e-05	2.32
5	1	7.2196e-04	5.7284e-06	0.0016	0.0024	7.2196e-06	3.32
6	1	0.0025	7.1011e-05	0.0055	0.0084	2.5175e-05	1.31
7	1	0.0026	7.4721e-05	0.0055	0.0086	2.5515e-05	2.03
8	1	8.2778e-04	7.9809e-06	0.0018	0.0028	8.2778e-06	1.48
9	1	7.3968e-04	6.2360e-06	0.0016	0.0025	7.3968e-06	3.32
10	1	7.5442e-04	6.5662e-06	0.0016	0.0026	7.5442e-06	3.50
11	1	0.0018	3.8319e-05	0.0040	0.0062	1.8279e-05	2.11
12	1	5.1926e-04	3.0656e-06	0.0011	0.0018	5.1926e-06	2.17
13	1	3.2149e-04	1.1717e-06	6.9914e-04	0.0011	3.2149e-06	5.04
14	1	2.3034e-04	5.9791e-07	5.0093e-04	7.7324e-04	2.3034e-06	1.32
15	1	0.0012	1.7649e-05	0.0027	0.0042	1.2348e-05	3.02

Bold implies the best results.