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

Table 5

The proposed five inputs based on improved backpropagation neural network statistical performance analysis with various hidden neurons for rainfall forecasting.

Number of hidden neurons	Error qualifier
	R	MAPE	MSE	MAE	RMSE	MRE	Time (min)
1	0.98949	8.9725	657.7246	18.1105	25.6461	0.0897	1.21
2	0.1638	100	7.4967e+04	201.8431	273.8012	1.000	0.01
3	0.99991	0.7373	5.4600	1.4882	2.3367	0.0074	0.57
4	1	0.2696	0.6398	0.5441	0.7999	0.0027	1.59
5	1	0.0945	0.0864	0.1906	0.2939	9.4451e-04	4.15
6	1	0.0344	0.0143	0.0695	0.1194	3.4418e-04	1.54
7	1	0.0158	0.0042	0.0320	0.0645	1.5841e-04	1.24
8	1	0.0203	0.0077	0.0409	0.0880	2.0264e-04	1.36
9	1	0.0199	0.0079	0.0401	0.0869	1.9879e-04	1.49
10	1	0.0170	0.0066	0.0344	0.0815	1.7022e-04	1.52
11	1	0.0357	0.1512	0.0721	0.3889	3.5704e-04	1.13
12	1	0.0283	0.0842	0.0572	0.2902	2.8320e-04	3.47
13	1	0.0231	0.2452	0.0466	0.4956	2.3092e-04	3.10
14	1	0.0191	0.3369	0.0385	0.5804	1.9076e-04	5.40
15	1	0.0255	0.6127	0.0515	0.7828	2.5517e-04	8.31

Bold implies the best results.