from abc import ABC
from streamline.modeling.basemodel import BaseModel
from streamline.modeling.parameters import get_parameters
from sklearn.ensemble import RandomForestClassifier as RF
[docs]
class RandomForestClassifier(BaseModel, ABC):
model_name = "Random Forest"
small_name = "RF"
color = "blue"
def __init__(self, cv_folds=3, scoring_metric='balanced_accuracy',
metric_direction='maximize', random_state=None, cv=None, n_jobs=None):
super().__init__(RF, "Random Forest", cv_folds, scoring_metric, metric_direction, random_state, cv)
self.param_grid = get_parameters(self.model_name)
self.param_grid['random_state'] = [random_state, ]
self.small_name = "RF"
self.color = "blue"
self.n_jobs = n_jobs
[docs]
def objective(self, trial, params=None):
self.params = {'n_estimators': trial.suggest_int('n_estimators', self.param_grid['n_estimators'][0],
self.param_grid['n_estimators'][1]),
'criterion': trial.suggest_categorical('criterion', self.param_grid['criterion']),
'max_depth': trial.suggest_int('max_depth', self.param_grid['max_depth'][0],
self.param_grid['max_depth'][1]),
'min_samples_split': trial.suggest_int('min_samples_split',
self.param_grid['min_samples_split'][0],
self.param_grid['min_samples_split'][1]),
'min_samples_leaf': trial.suggest_int('min_samples_leaf', self.param_grid['min_samples_leaf'][0],
self.param_grid['min_samples_leaf'][1]),
'max_features': trial.suggest_categorical('max_features', self.param_grid['max_features']),
'bootstrap': trial.suggest_categorical('bootstrap', self.param_grid['bootstrap']),
'oob_score': trial.suggest_categorical('oob_score', self.param_grid['oob_score']),
'class_weight': trial.suggest_categorical('class_weight', self.param_grid['class_weight']),
'random_state': trial.suggest_categorical('random_state', self.param_grid['random_state'])}
mean_cv_score = self.hyper_eval()
return mean_cv_score