Table 6 Optimal AI training parameters for hourly SR task.
Location | Model | No. of hidden layers, [No. of neurons in each hidden layer] | Batch size | Epoch |
|---|---|---|---|---|
Algeria GSR | ANN | 3, [100, 100, 50] | 512 | 100 |
CNN-ANN | 7, [64, 64, 32, (), 100, 100, 50] | 512 | 100 | |
CNN-LSTM-ANN | 6, [32, 32, 32, 50, 50, 25] | 512 | 100 | |
CNN | 2, [150, 100] | 512 | 100 | |
LSTM | 2, [150, 100] | 512 | 15 | |
Nigeria DSR | ANN | 2, [100, 50] | 512 | 50 |
CNN-ANN | 3, [64, (), 50] | 512 | 30 | |
CNN-LSTM-ANN | 3, [32, 32, 50] | 512 | 30 | |
CNN | 2, [150, 100] | 512 | 50 | |
LSTM | 1, [100] | 512 | 20 | |
Central African Republic GSR | ANN | 3, [200, 200, 100] | 512 | 30 |
CNN-ANN | 7, [32, 64, 32, (), 32, 100, 32] | 512 | 10 | |
CNN-LSTM-ANN | 6, [32, 16, 32, 25, 50, 25] | 512 | 10 | |
CNN | 2, [150, 100] | 512 | 30 | |
LSTM | 1, [150] | 512 | 10 | |
Egypt GSR | ANN | 3, [200, 200, 100] | 128 | 7 |
CNN-ANN | 7, [32, 64, 32, (), 32, 100, 32] | 512 | 10 | |
CNN-LSTM-ANN | 6, [32, 16, 32, 25, 50, 25] | 512 | 10 | |
CNN | 2, [150, 100] | 512 | 30 | |
LSTM | 2, [150, 100] | 512 | 50 | |
South Africa GSR | ANN | 2, [100, 50] | 512 | 20 |
CNN-ANN | 3, [64 (), 32] | 512 | 20 | |
CNN-LSTM-ANN | 6, [32, 16, 32, 25, 50, 25] | 512 | 10 | |
CNN | 2, [150, 100] | 512 | 20 | |
LSTM | 2, [50, 50] | 512 | 50 |
