Module automl_alex.models.model_xgboost

Expand source code 
from automl_alex._base import ModelBase
import xgboost as xgb
import numpy as np


class XGBoost(ModelBase):
    """
    Args:
        params (dict or None): parameters for model.
            If None default params are fetched.
    """
    __name__ = 'XGBoost'

    def _init_default_model_param(self,):
        """
        Default model_param
        """
        model_param = {
                        'verbosity': 0,
                        'n_estimators': 300,
                        'random_state': self._random_state,
                        }
        return(model_param)

    def _init_model(self, model_param=None):
        """
        sets new model,
        Args:
            params: : parameters for model.
        """
        if self._type_of_estimator == 'classifier':
            model = xgb.XGBClassifier(**model_param)
        elif self._type_of_estimator == 'regression':
            model = xgb.XGBRegressor(**model_param)
        return(model)

    #@staticmethod
    def get_model_opt_params(self, trial, opt_lvl):
        """
        Return:
            dict of DistributionWrappers
        """
        model_param = self._init_default_model_param()
        ################################# LVL 1 ########################################
        if opt_lvl == 1:
            model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 2, 8,)
            model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 7)
        
        ################################# LVL 2 ########################################
        if opt_lvl == 2:
            model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 2, 12,)
            model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 10)
        
        if opt_lvl >= 2:
            model_param['learning_rate'] = trial.suggest_int('xgb_learning_rate', 1, 100)/1000
            model_param['subsample'] = trial.suggest_discrete_uniform('xgb_subsample', 0.1, 1., 0.1)
            model_param['colsample_bytree'] = trial.suggest_discrete_uniform('xgb_colsample_bytree', 0.1, 1., 0.1)
            
        ################################# LVL 3 ########################################
        if opt_lvl >= 3:
            model_param['booster'] = trial.suggest_categorical('xgb_booster', ['gbtree', 'dart', 'gblinear'])
            
            if model_param['booster'] == 'dart' or model_param['booster'] == 'gbtree':
                model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 100)
                model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 1, 20)
                model_param['gamma'] = trial.suggest_loguniform('xgb_gamma', 1e-6, 1.0)
                model_param['grow_policy'] = trial.suggest_categorical('xgb_grow_policy', ['depthwise', 'lossguide'])
            
            if model_param['booster'] == 'dart':
                model_param['early_stopping_rounds'] = 0
                model_param['sample_type'] = trial.suggest_categorical('xgb_sample_type', ['uniform', 'weighted'])
                model_param['normalize_type'] = trial.suggest_categorical('xgb_normalize_type', ['tree', 'forest'])
                model_param['rate_drop'] = trial.suggest_loguniform('xgb_rate_drop', 1e-8, 1.0)
                model_param['skip_drop'] = trial.suggest_loguniform('xgb_skip_drop', 1e-8, 1.0)

            model_param['n_estimators'] = trial.suggest_int('xgb_n_estimators', 1, 10,)*100

        ################################# LVL 4 ########################################
        if opt_lvl >= 4:
            if self._type_of_estimator == 'regression':
                model_param['objective'] = trial.suggest_categorical('xgb_objective', 
                    [
                    'reg:squarederror',
                    'reg:squaredlogerror',
                    'reg:logistic',
                    ])
            
            model_param['lambda'] = trial.suggest_loguniform('xg_lambda', 1e-8, 1.0)
            model_param['regalpha_alpha'] = trial.suggest_loguniform('XG_alpha', 1e-8, 1.0)
        
        ################################# Other ########################################
        return(model_param)

    
    def _is_model_start_opt_params(self,):
        return(False)


    def fit(self, X_train=None, y_train=None, cat_features=None):
        """
        Args:
            X (pd.DataFrame, shape (n_samples, n_features)): the input data
            y (pd.DataFrame, shape (n_samples, ) or (n_samples, n_outputs)): the target data
        Return:
            self
        """
        params = self.model_param.copy()

        self.model = self._init_model(model_param=params)

        self.model.fit(
            X_train, 
            y_train,
            verbose=False,
            )
        return self

    def predict(self, X=None):
        """
        Args:
            X (np.array, shape (n_samples, n_features)): the input data
        Return:
            np.array, shape (n_samples, n_classes)
        """
        if self.model is None:
            raise Exception("No fit models")

        if self._type_of_estimator == 'classifier':
            predicts = np.round(self.model.predict(X),0)

        elif self._type_of_estimator == 'regression':
            predicts = self.model.predict(X)
        return predicts

    def is_possible_predict_proba(self):
        """
        Return:
            bool, whether model can predict proba
        """
        return True

    def predict_proba(self, X):
        """
        Args:
            X (np.array, shape (n_samples, n_features)): the input data
        Return:
            np.array, shape (n_samples, n_classes)
        """
        if self.model is None:
            raise Exception("No fit models")

        if not self.is_possible_predict_proba(): 
            raise Exception("Model cannot predict probability distribution")

        predicts = self.model.predict(X)
        return predicts

    def _is_possible_feature_importance(self):
        """
        Return:
            bool, whether model can predict proba
        """
        return False


class XGBoostClassifier(XGBoost):
    _type_of_estimator='classifier'


class XGBoostRegressor(XGBoost):
    _type_of_estimator='regression'

Classes

class XGBoost (model_param=None, type_of_estimator=None, gpu=False, verbose=None, random_state=42)

Args

params : dict or None
parameters for model. If None default params are fetched.
Expand source code 
class XGBoost(ModelBase):
    """
    Args:
        params (dict or None): parameters for model.
            If None default params are fetched.
    """
    __name__ = 'XGBoost'

    def _init_default_model_param(self,):
        """
        Default model_param
        """
        model_param = {
                        'verbosity': 0,
                        'n_estimators': 300,
                        'random_state': self._random_state,
                        }
        return(model_param)

    def _init_model(self, model_param=None):
        """
        sets new model,
        Args:
            params: : parameters for model.
        """
        if self._type_of_estimator == 'classifier':
            model = xgb.XGBClassifier(**model_param)
        elif self._type_of_estimator == 'regression':
            model = xgb.XGBRegressor(**model_param)
        return(model)

    #@staticmethod
    def get_model_opt_params(self, trial, opt_lvl):
        """
        Return:
            dict of DistributionWrappers
        """
        model_param = self._init_default_model_param()
        ################################# LVL 1 ########################################
        if opt_lvl == 1:
            model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 2, 8,)
            model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 7)
        
        ################################# LVL 2 ########################################
        if opt_lvl == 2:
            model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 2, 12,)
            model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 10)
        
        if opt_lvl >= 2:
            model_param['learning_rate'] = trial.suggest_int('xgb_learning_rate', 1, 100)/1000
            model_param['subsample'] = trial.suggest_discrete_uniform('xgb_subsample', 0.1, 1., 0.1)
            model_param['colsample_bytree'] = trial.suggest_discrete_uniform('xgb_colsample_bytree', 0.1, 1., 0.1)
            
        ################################# LVL 3 ########################################
        if opt_lvl >= 3:
            model_param['booster'] = trial.suggest_categorical('xgb_booster', ['gbtree', 'dart', 'gblinear'])
            
            if model_param['booster'] == 'dart' or model_param['booster'] == 'gbtree':
                model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 100)
                model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 1, 20)
                model_param['gamma'] = trial.suggest_loguniform('xgb_gamma', 1e-6, 1.0)
                model_param['grow_policy'] = trial.suggest_categorical('xgb_grow_policy', ['depthwise', 'lossguide'])
            
            if model_param['booster'] == 'dart':
                model_param['early_stopping_rounds'] = 0
                model_param['sample_type'] = trial.suggest_categorical('xgb_sample_type', ['uniform', 'weighted'])
                model_param['normalize_type'] = trial.suggest_categorical('xgb_normalize_type', ['tree', 'forest'])
                model_param['rate_drop'] = trial.suggest_loguniform('xgb_rate_drop', 1e-8, 1.0)
                model_param['skip_drop'] = trial.suggest_loguniform('xgb_skip_drop', 1e-8, 1.0)

            model_param['n_estimators'] = trial.suggest_int('xgb_n_estimators', 1, 10,)*100

        ################################# LVL 4 ########################################
        if opt_lvl >= 4:
            if self._type_of_estimator == 'regression':
                model_param['objective'] = trial.suggest_categorical('xgb_objective', 
                    [
                    'reg:squarederror',
                    'reg:squaredlogerror',
                    'reg:logistic',
                    ])
            
            model_param['lambda'] = trial.suggest_loguniform('xg_lambda', 1e-8, 1.0)
            model_param['regalpha_alpha'] = trial.suggest_loguniform('XG_alpha', 1e-8, 1.0)
        
        ################################# Other ########################################
        return(model_param)

    
    def _is_model_start_opt_params(self,):
        return(False)


    def fit(self, X_train=None, y_train=None, cat_features=None):
        """
        Args:
            X (pd.DataFrame, shape (n_samples, n_features)): the input data
            y (pd.DataFrame, shape (n_samples, ) or (n_samples, n_outputs)): the target data
        Return:
            self
        """
        params = self.model_param.copy()

        self.model = self._init_model(model_param=params)

        self.model.fit(
            X_train, 
            y_train,
            verbose=False,
            )
        return self

    def predict(self, X=None):
        """
        Args:
            X (np.array, shape (n_samples, n_features)): the input data
        Return:
            np.array, shape (n_samples, n_classes)
        """
        if self.model is None:
            raise Exception("No fit models")

        if self._type_of_estimator == 'classifier':
            predicts = np.round(self.model.predict(X),0)

        elif self._type_of_estimator == 'regression':
            predicts = self.model.predict(X)
        return predicts

    def is_possible_predict_proba(self):
        """
        Return:
            bool, whether model can predict proba
        """
        return True

    def predict_proba(self, X):
        """
        Args:
            X (np.array, shape (n_samples, n_features)): the input data
        Return:
            np.array, shape (n_samples, n_classes)
        """
        if self.model is None:
            raise Exception("No fit models")

        if not self.is_possible_predict_proba(): 
            raise Exception("Model cannot predict probability distribution")

        predicts = self.model.predict(X)
        return predicts

    def _is_possible_feature_importance(self):
        """
        Return:
            bool, whether model can predict proba
        """
        return False

Ancestors

  • automl_alex._base.ModelBase

Subclasses

Methods

def fit(self, X_train=None, y_train=None, cat_features=None)

Args

X (pd.DataFrame, shape (n_samples, n_features)): the input data y (pd.DataFrame, shape (n_samples, ) or (n_samples, n_outputs)): the target data

Return

self

Expand source code 
def fit(self, X_train=None, y_train=None, cat_features=None):
    """
    Args:
        X (pd.DataFrame, shape (n_samples, n_features)): the input data
        y (pd.DataFrame, shape (n_samples, ) or (n_samples, n_outputs)): the target data
    Return:
        self
    """
    params = self.model_param.copy()

    self.model = self._init_model(model_param=params)

    self.model.fit(
        X_train, 
        y_train,
        verbose=False,
        )
    return self
def get_model_opt_params(self, trial, opt_lvl)

Return

dict of DistributionWrappers

Expand source code 
def get_model_opt_params(self, trial, opt_lvl):
    """
    Return:
        dict of DistributionWrappers
    """
    model_param = self._init_default_model_param()
    ################################# LVL 1 ########################################
    if opt_lvl == 1:
        model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 2, 8,)
        model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 7)
    
    ################################# LVL 2 ########################################
    if opt_lvl == 2:
        model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 2, 12,)
        model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 10)
    
    if opt_lvl >= 2:
        model_param['learning_rate'] = trial.suggest_int('xgb_learning_rate', 1, 100)/1000
        model_param['subsample'] = trial.suggest_discrete_uniform('xgb_subsample', 0.1, 1., 0.1)
        model_param['colsample_bytree'] = trial.suggest_discrete_uniform('xgb_colsample_bytree', 0.1, 1., 0.1)
        
    ################################# LVL 3 ########################################
    if opt_lvl >= 3:
        model_param['booster'] = trial.suggest_categorical('xgb_booster', ['gbtree', 'dart', 'gblinear'])
        
        if model_param['booster'] == 'dart' or model_param['booster'] == 'gbtree':
            model_param['min_child_weight'] = trial.suggest_int('xgb_min_child_weight', 2, 100)
            model_param['max_depth'] = trial.suggest_int('xgb_max_depth', 1, 20)
            model_param['gamma'] = trial.suggest_loguniform('xgb_gamma', 1e-6, 1.0)
            model_param['grow_policy'] = trial.suggest_categorical('xgb_grow_policy', ['depthwise', 'lossguide'])
        
        if model_param['booster'] == 'dart':
            model_param['early_stopping_rounds'] = 0
            model_param['sample_type'] = trial.suggest_categorical('xgb_sample_type', ['uniform', 'weighted'])
            model_param['normalize_type'] = trial.suggest_categorical('xgb_normalize_type', ['tree', 'forest'])
            model_param['rate_drop'] = trial.suggest_loguniform('xgb_rate_drop', 1e-8, 1.0)
            model_param['skip_drop'] = trial.suggest_loguniform('xgb_skip_drop', 1e-8, 1.0)

        model_param['n_estimators'] = trial.suggest_int('xgb_n_estimators', 1, 10,)*100

    ################################# LVL 4 ########################################
    if opt_lvl >= 4:
        if self._type_of_estimator == 'regression':
            model_param['objective'] = trial.suggest_categorical('xgb_objective', 
                [
                'reg:squarederror',
                'reg:squaredlogerror',
                'reg:logistic',
                ])
        
        model_param['lambda'] = trial.suggest_loguniform('xg_lambda', 1e-8, 1.0)
        model_param['regalpha_alpha'] = trial.suggest_loguniform('XG_alpha', 1e-8, 1.0)
    
    ################################# Other ########################################
    return(model_param)
def is_possible_predict_proba(self)

Return

bool, whether model can predict proba

Expand source code 
def is_possible_predict_proba(self):
    """
    Return:
        bool, whether model can predict proba
    """
    return True
def predict(self, X=None)

Args

X (np.array, shape (n_samples, n_features)): the input data

Return

np.array, shape (n_samples, n_classes)

Expand source code 
def predict(self, X=None):
    """
    Args:
        X (np.array, shape (n_samples, n_features)): the input data
    Return:
        np.array, shape (n_samples, n_classes)
    """
    if self.model is None:
        raise Exception("No fit models")

    if self._type_of_estimator == 'classifier':
        predicts = np.round(self.model.predict(X),0)

    elif self._type_of_estimator == 'regression':
        predicts = self.model.predict(X)
    return predicts
def predict_proba(self, X)

Args

X (np.array, shape (n_samples, n_features)): the input data

Return

np.array, shape (n_samples, n_classes)

Expand source code 
def predict_proba(self, X):
    """
    Args:
        X (np.array, shape (n_samples, n_features)): the input data
    Return:
        np.array, shape (n_samples, n_classes)
    """
    if self.model is None:
        raise Exception("No fit models")

    if not self.is_possible_predict_proba(): 
        raise Exception("Model cannot predict probability distribution")

    predicts = self.model.predict(X)
    return predicts
class XGBoostClassifier (model_param=None, type_of_estimator=None, gpu=False, verbose=None, random_state=42)

Args

params : dict or None
parameters for model. If None default params are fetched.
Expand source code 
class XGBoostClassifier(XGBoost):
    _type_of_estimator='classifier'

Ancestors

  • XGBoost
  • automl_alex._base.ModelBase

Inherited members

class XGBoostRegressor (model_param=None, type_of_estimator=None, gpu=False, verbose=None, random_state=42)

Args

params : dict or None
parameters for model. If None default params are fetched.
Expand source code 
class XGBoostRegressor(XGBoost):
    _type_of_estimator='regression'

Ancestors

  • XGBoost
  • automl_alex._base.ModelBase

Inherited members