Forecasting Different Types of Droughts Simultaneously Using Multivariate Standardized Precipitation Index (MSPI), MLP Neural Network, and Imperialistic Competitive Algorithm (ICA)
Table 4
Evaluating the predictions of the models MLP and MLP-ICA.
Station
Time window
Model
RMSE
MAE
R
Train
Test
Train
Test
Train
Test
Hamedan
MSPI3–6
MLP
0.42
0.63
0.31
0.46
0.91
0.78
MLP-ICA
0.40
0.60
0.29
0.46
0.92
0.81
MSPI6–12
MLP
0.31
0.50
0.22
0.38
0.95
0.87
MLP-ICA
0.29
0.40
0.21
0.28
0.95
0.92
MSPI3–12
MLP
0.30
0.54
0.21
0.38
0.95
0.84
MLP-ICA
0.30
0.53
0.22
0.38
0.95
0.86
MSPI12–24
MLP
0.23
0.36
0.17
0.26
0.98
0.89
MLP-ICA
0.20
0.27
0.14
0.20
0.98
0.94
MSPI24–48
MLP
0.19
0.32
0.15
0.26
0.98
0.91
MLP-ICA
0.14
0.20
0.10
0.15
0.99
0.95
Kermanshah
MSPI3–6
MLP
0.42
0.73
0.29
0.57
0.91
0.73
MLP-ICA
0.42
0.64
0.31
0.47
0.90
0.80
MSPI6–12
MLP
0.42
0.71
0.29
0.53
0.92
0.67
MLP-ICA
0.32
0.59
0.23
0.41
0.95
0.79
MSPI3–12
MLP
0.40
0.80
0.30
0.60
0.92
0.60
MLP-ICA
0.40
0.63
0.28
0.44
0.92
0.79
MSPI12–24
MLP
0.24
0.47
0.18
0.37
0.98
0.79
MLP-ICA
0.16
0.31
0.12
0.24
0.99
0.92
MSPI24–48
MLP
0.18
0.30
0.14
0.25
0.99
0.78
MLP-ICA
0.12
0.20
0.09
0.15
0.99
0.92
The bold numbers are the best predictions of each station.
