Lgbm optimization using hyperopt

lgbm_hyperopt_optimization.py

from hyperopt import hp, fmin, tpe, Trials, STATUS_OK
lgb_reg_params = {
    'learning_rate':    hp.uniform('learning_rate',0.1,1),
    'max_depth':        hp.choice('max_depth',        np.arange(2, 100, 1, dtype=int)),
    'min_child_weight': hp.choice('min_child_weight', np.arange(1, 50, 1, dtype=int)),
    'colsample_bytree': hp.uniform('colsample_bytree',0.4,1),
    'subsample':        hp.uniform('subsample', 0.6, 1),
    'num_leaves':       hp.choice('num_leaves',       np.arange(1, 200, 1, dtype=int)),
    'min_split_gain':   hp.uniform('min_split_gain', 0, 1),
    'reg_alpha':        hp.uniform('reg_alpha',0,1),
    'reg_lambda':       hp.uniform('reg_lambda',0,1),
    'n_estimators':     5
}
def f(params):
  lgbm = lgb.LGBMRegressor(n_jobs=-1,early_stopping_rounds=None,**params)
  score = cross_val_score(lgbm, X_train, y_train, cv=2,scoring=rmsle,n_jobs=-1).mean()
  return score

trials = Trials()
result = fmin(
    fn=f,                           # objective function
    space=lgb_reg_params,   # parameter space
    algo=tpe.suggest,               # surrogate algorithm
    max_evals=50,                  # no. of evaluations
    trials=trials                   # trials object that keeps track of the sample results (optional)
)
print(result)

I suggest you shap-hypetune to industrialize parameter tuning (and also feature selection) with LGBM and hyperopt (https://github.com/cerlymarco/shap-hypetune)

Thanks for the suggestion. I have bookmarked the repo