Bayesian Optimization LightGBM 적용

import numpy as np

from xgboost import XGBClassifier

from bayes_opt import BayesianOptimization

from sklearn.model_selection import cross_val_score

pbounds = {

'learning_rate': (0.01, 0.5),

'n_estimators': (100, 1000),

'max_depth': (3, 10),

'min_child_weight': (0, 10),

'subsample': (0.5, 1.0),

'colsample_bytree': (0.5, 1.0)

# 'reg_lambda': (0, 1000),

# 'reg_alpha': (0, 1.0)

}

def lgbm_hyper_param(learning_rate, n_estimators, max_depth, min_child_weight, subsample, colsample_bytree):

max_depth = int(max_depth)

n_estimators = int(n_estimators)

clf = LGBMClassifier(

max_depth=max_depth,

min_child_weight=min_child_weight,

learning_rate=learning_rate,

n_estimators=n_estimators,

subsample=subsample,

colsample_bytree=colsample_bytree,

random_state=1

# reg_lambda=reg_lambda,

# reg_alpha=reg_alpha

)

return np.mean(cross_val_score(clf, train_importance, train_answer, cv=5, scoring='accuracy')) # cv 도 숫자로 작성하여, 내부적으로 (Stratified)KFold 사용함

optimizer = BayesianOptimization( f=lgbm_hyper_param, pbounds=pbounds, verbose=1, random_state=1)

optimizer.maximize(init_points=10, n_iter=100, acq='ei', xi=0.01)

에서

| iter | target | colsam... | learni... | max_depth | min_ch... | n_esti... | subsample | ------------------------------------------------------------------------------------------------- [LightGBM] [Info] Number of positive: 273, number of negative: 439 [LightGBM] [Info] Auto-choosing row-wise multi-threading, the overhead of testing was 0.000185 seconds. You can set force_row_wise=true to remove the overhead. And if memory is not enough, you can set force_col_wise=true. [LightGBM] [Info] Total Bins 90 [LightGBM] [Info] Number of data points in the train set: 712, number of used features: 45 [LightGBM] [Info] [binary:BoostFromScore]: pavg=0.383427 -> initscore=-0.475028 [LightGBM] [Info] Start training from score -0.475028 [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf [LightGBM] [Warning] No further splits with positive gain, best gain: -inf

같은 오류가 발생합니다...