Table 7 Results of hyperparameter optimization for LightGBM
From: Comparison of factors influencing landslide risk near a forest road in Chungju-si, South Korea
Train and test data ratio | 9:1 | 8:2 | 7:3 | 6:4 | 5:5 |
|---|---|---|---|---|---|
learning_rate | 0.01 | 0.01 | 0.01 | 0.01 | 0.01 |
n_estimators | 1500 | 1500 | 1500 | 500 | 500 |
max_depth | 2 | 2 | 2 | 2 | 2 |
gamma | – | – | – | – | – |
colsample_bytree | – | – | – | – | – |
max_delta_step | 0 | 0 | 0 | 0 | 0 |
min_child_weight | 0 | 0 | 0 | 0 | 0 |
reg_alpha | 0 | 2 | 2 | 3 | 0 |
reg_lambda | 0 | 0 | 0 | 0 | 0 |
subsample | 1 | 1 | 1 | 1 | 1 |
scale_pos_weight | 1 | 1 | 1 | 2 | 1 |