Hyperparameter optimization of different methods and the prediction performance
| LightGBM | Learning rate | Num leaves | Max depth | Min data in leaf | Feature fraction | RMSE | Pearson | Error |
|---|---|---|---|---|---|---|---|---|
| NGG | ||||||||
| Single parameter | 0.1 | 31 | −1 | 20 | 1 | 11.2 | 0.21 | 7.34 |
| Grid search | 0.05 | 15 | 3 | 30 | 0.6 | 10.67 | 0.21 | 6.94 |
| Sparrow search | 0.02 | 22 | −1 | 10 | 0.8 | 9.5 | 0.23 | 5.34 |
| AGE | ||||||||
| Single parameter | 0.1 | 31 | −1 | 20 | 1 | 4.03 | 0.45 | 3.12 |
| Grid search | 0.1 | 18 | 1 | 60 | 0.9 | 3.26 | 0.49 | 2.75 |
| Sparrow search | 0.08 | 12 | −1 | 35 | 1 | 2.94 | 0.50 | 2.01 |
| TW | ||||||||
| Single parameter | 0.1 | 31 | −1 | 20 | 1 | 9.2 | 0.17 | 7.45 |
| Grid search | 0.03 | 20 | 1 | 50 | 0.8 | 6.18 | 0.24 | 4.12 |
| Sparrow search | 0.05 | 26 | −1 | 80 | 0.6 | 4.61 | 0.27 | 3.89 |
| CatBoost | Learning rate | Depth | L2 Leaf reg | Iterations | Border count | RMSE | Pearson | Error |
|---|---|---|---|---|---|---|---|---|
| NGG | ||||||||
| Single parameter | 0.03 | 6 | 3 | 1000 | 254 | 10.32 | 0.22 | 6.21 |
| Grid search | 0.05 | 8 | 5 | 1500 | 200 | 10.13 | 0.23 | 6.06 |
| Sparrow search | 0.04 | 7 | 3 | 1200 | 86 | 9.97 | 0.24 | 5.9 |
| AGE | ||||||||
| Single parameter | 0.03 | 6 | 3 | 1000 | 254 | 6.86 | 0.38 | 5.11 |
| Grid search | 0.1 | 10 | 6 | 1300 | 64 | 5.05 | 0.43 | 4.22 |
| Sparrow search | 0.06 | 7 | 5 | 1100 | 64 | 4.78 | 0.44 | 3.76 |
| TW | ||||||||
| Single parameter | 0.03 | 6 | 3 | 1000 | 254 | 8.63 | 0.18 | 7.08 |
| Grid search | 0.06 | 10 | 5 | 700 | 64 | 5.48 | 0.26 | 3.75 |
| Sparrow search | 0.05 | 8 | 5 | 800 | 60 | 3.99 | 0.28 | 3.23 |