Table 3 Selected parameters for SVR and LSTM in the combination approaches using a grid search in the validation set for 1 day ahead SST forecasting.
From: Hybrid systems using residual modeling for sea surface temperature forecasting
Time series | Model | Parameters | Combination approaches | ||||
|---|---|---|---|---|---|---|---|
Perturbative22 | NoLiC23 | ||||||
\({\mathrm{M}}_0\) | \({\mathrm{M}}_1\) | \({\mathrm{M}}_2\) | \({\mathrm{M}}_3\) | \({\mathrm{M}}_{\mathrm{C}}\) | |||
S1 | SVR | Gamma | 1 | 1 | 0.001 | 1 | 1 |
Cost | 1 | 0.1 | 1 | 1 | 100 | ||
Tolerance | 0.001 | 0.001 | 0.01 | 0.001 | 0.01 | ||
Inputs | 2 | 2 | 1 | 1 | 2 | ||
LSTM | Units in hidden layer | 2 | 5 | 5 | – | 10 | |
Inputs | 1 | 2 | 2 | – | 2 | ||
S2 | SVR | Gamma | 0.001 | 1 | 1 | 0.001 | 1 |
Cost | 100 | 1 | 0.1 | 100 | 1 | ||
Tolerance | 0.001 | 0.01 | 0.01 | 0.001 | 0.001 | ||
Inputs | 23 | 2 | 2 | 2 | 2 | ||
LSTM | Units in hidden layer | 5 | 10 | 5 | – | 5 | |
Inputs | 5 | 2 | 2 | – | 2 | ||
S3 | SVR | Gamma | 1 | 0.001 | 1 | 1 | 1 |
Cost | 1 | 100 | 1 | 100 | 100 | ||
Tolerance | 0.001 | 0.01 | 0.01 | 0.01 | 0.01 | ||
Inputs | 3 | 2 | 2 | 2 | 2 | ||
LSTM | Units in hidden layer | 5 | 5 | 5 | – | 10 | |
Inputs | 5 | 2 | 2 | – | 2 | ||
