Table 3 LightGBM set of optimized hyperparameters with search boundaries.
Parameter | Lower constraint | Upper constraint |
|---|---|---|
Learning rate | 0.001 | 0.100 |
Number of leaves | 24 | 80 |
Feature fraction | 0.1 | 0.9 |
Bagging fraction | 0.8 | 1 |
Max depth | 5 | 30 |
Max bin | 20 | 90 |
Minimum data in leaf | 20 | 80 |
Minimum sum hessian in leaf | 0 | 100 |
Subsample | 0.01 | 1 |