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

Table 6

The proposed five input-based Elman neural network statistical performance analyses with various hidden neurons for temperature forecasting.

Number of hidden neurons	Error qualifier
	MAPE	MSE	MAE	RMSE	MRE	Time (sec)
1	0.6612	0.1299	0.1488	0.3605	0.0066	22
2	1.2701	0.4218	0.2859	0.6494	0.0127	31
3	0.8163	0.1949	0.1837	0.4415	0.0082	29
4	0.6202	0.0684	0.1396	0.2615	0.0062	33
5	1.0752	0.3555	0.2420	0.5962	0.0108	34
6	0.4970	0.0821	0.1119	0.2866	0.0050	33
7	0.1069	0.0035	0.0241	0.0590	0.0011	44
8	0.1013	0.0015	0.0228	0.0384	0.0010	35
9	0.4501	0.0671	0.1013	0.2591	0.0045	42
10	0.9629	0.2676	0.2167	0.5173	0.0096	45
11	0.1023	0.0011	0.0230	0.0332	0.0010	22
12	0.3122	0.0319	0.0703	0.1787	0.0031	46
13	0.1715	0.0025	0.0386	0.0502	0.0017	39
14	0.4440	0.0473	0.0999	0.2175	0.0044	49
15	0.8641	0.2005	0.1945	0.4478	0.0086	51

Bold implies the best results.