Skip to main content

Table 4 Changed hyperparameter settings

From: Evaluating classifier performance with highly imbalanced Big Data

Classifier/

Hyperparameter

Setting

Description

CatBoost

task_type

GPU

Use GPU(s) for execution

devices

0

GPU device ID

max_ctr_complexity

1

Maximum number of features to combine

max_depth

16

Maximum tree depth

n_estimators

100

Number of trees in ensemble

XGBoost

tree_method

gpu_hist

Use GPU(s) for execution

gpu_id

0

GPU device ID

n_estimators

100

Number of trees in ensemble

max_depth

24

Maximum tree depth

LightGBM

num_threads

64

Maximum number of execution threads

Random Forest

max_depth

32

Maximum tree depth

n_streams

8

Number of simultaneous GPU streams active during the fit phase