Module automl_alex.models.model_catboost

Expand source code 
from automl_alex._base import ModelBase
import catboost
from catboost import Pool
import numpy as np
import pandas as pd


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


    def _init_default_model_param(self, model_param=None):
        """
        Default model_param
        """
        if model_param is None:
            model_param = {'verbose': 0, 
                                'task_type': 'GPU' if self._gpu else 'CPU',
                                'random_seed': 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 = catboost.CatBoostClassifier(**model_param)
        elif self._type_of_estimator == 'regression':
            model = catboost.CatBoostRegressor(**model_param)
        return(model)


    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
            cat_features (list)
        Return:
            self
        """
        y_train = self.y_format(y_train)

        if cat_features is not None:
            cat_dims = [X_train.columns.get_loc(i) for i in cat_features[:]]
            train_pool = Pool(X_train, label=y_train, cat_features=cat_dims)
        else:
            train_pool = Pool(X_train, label=y_train,)

        params = self.model_param.copy()
        self.model = self._init_model(model_param=params)
        self.model.fit(train_pool, verbose=False, plot=False,)
        train_pool=None
        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=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 not self.is_possible_predict_proba(): 
            raise Exception("Model cannot predict probability distribution")
        return self.model.predict_proba(X)[:, 1]


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


    def get_feature_importance(self, X,):
        """
        Return:
            list feature_importance
        """
        if not self._is_possible_feature_importance(): 
            raise Exception("Model cannot get feature_importance")

        fe_lst = self.model.get_feature_importance()
        return (pd.DataFrame(fe_lst, index=X.columns, columns=['value']))


    #@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['min_child_samples'] = trial.suggest_int('cb_min_child_samples', 1, 100)
            model_param['depth'] = trial.suggest_int('cb_depth', 4, 10)
            

        ################################# LVL 2 ########################################
        if opt_lvl >= 2:
            model_param['bagging_temperature'] = trial.suggest_int('cb_bagging_temperature', 0, 10,)
            model_param['subsample'] = trial.suggest_discrete_uniform('cb_subsample', 0.1, 1.0, 0.1)
        
        ################################# LVL 3 ########################################
        if opt_lvl >= 3:
            if self._type_of_estimator == 'classifier':
                model_param['objective'] = trial.suggest_categorical('cb_objective', 
                        [
                        'Logloss', 
                        'CrossEntropy',
                        ])

            elif self._type_of_estimator == 'regression':
                model_param['objective'] = trial.suggest_categorical('cb_objective', 
                    [
                    'MAE',
                    'MAPE',
                    'Quantile',
                    'RMSE',
                    ])
        

        ################################# LVL 4 ########################################
        if opt_lvl >= 4:
            model_param['depth'] = trial.suggest_int('cb_depth', 2, 16)
            model_param['l2_leaf_reg'] = trial.suggest_loguniform('cb_l2_leaf_reg', 1e-8, .1)
            model_param['learning_rate'] = trial.suggest_int('cb_learning_rate', 1, 100)/1000
            model_param['iterations'] = trial.suggest_int('cb_iterations', 1, 10)*100

        ################################# Other ########################################
        return(model_param)

    def _is_model_start_opt_params(self,):
        return(True)


    def get_model_start_opt_params(self,):
        dafault_params = {
            "cb_depth": 6,
            "cb_min_child_samples": 1,
            "cb_learning_rate": 0.03,
            }
        return(dafault_params)


class CatBoostClassifier(CatBoost):
    _type_of_estimator='classifier'
    __name__ = 'CatBoostClassifier'


class CatBoostRegressor(CatBoost):
    _type_of_estimator='regression'
    __name__ = 'CatBoostRegressor'

Classes

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


    def _init_default_model_param(self, model_param=None):
        """
        Default model_param
        """
        if model_param is None:
            model_param = {'verbose': 0, 
                                'task_type': 'GPU' if self._gpu else 'CPU',
                                'random_seed': 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 = catboost.CatBoostClassifier(**model_param)
        elif self._type_of_estimator == 'regression':
            model = catboost.CatBoostRegressor(**model_param)
        return(model)


    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
            cat_features (list)
        Return:
            self
        """
        y_train = self.y_format(y_train)

        if cat_features is not None:
            cat_dims = [X_train.columns.get_loc(i) for i in cat_features[:]]
            train_pool = Pool(X_train, label=y_train, cat_features=cat_dims)
        else:
            train_pool = Pool(X_train, label=y_train,)

        params = self.model_param.copy()
        self.model = self._init_model(model_param=params)
        self.model.fit(train_pool, verbose=False, plot=False,)
        train_pool=None
        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=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 not self.is_possible_predict_proba(): 
            raise Exception("Model cannot predict probability distribution")
        return self.model.predict_proba(X)[:, 1]


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


    def get_feature_importance(self, X,):
        """
        Return:
            list feature_importance
        """
        if not self._is_possible_feature_importance(): 
            raise Exception("Model cannot get feature_importance")

        fe_lst = self.model.get_feature_importance()
        return (pd.DataFrame(fe_lst, index=X.columns, columns=['value']))


    #@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['min_child_samples'] = trial.suggest_int('cb_min_child_samples', 1, 100)
            model_param['depth'] = trial.suggest_int('cb_depth', 4, 10)
            

        ################################# LVL 2 ########################################
        if opt_lvl >= 2:
            model_param['bagging_temperature'] = trial.suggest_int('cb_bagging_temperature', 0, 10,)
            model_param['subsample'] = trial.suggest_discrete_uniform('cb_subsample', 0.1, 1.0, 0.1)
        
        ################################# LVL 3 ########################################
        if opt_lvl >= 3:
            if self._type_of_estimator == 'classifier':
                model_param['objective'] = trial.suggest_categorical('cb_objective', 
                        [
                        'Logloss', 
                        'CrossEntropy',
                        ])

            elif self._type_of_estimator == 'regression':
                model_param['objective'] = trial.suggest_categorical('cb_objective', 
                    [
                    'MAE',
                    'MAPE',
                    'Quantile',
                    'RMSE',
                    ])
        

        ################################# LVL 4 ########################################
        if opt_lvl >= 4:
            model_param['depth'] = trial.suggest_int('cb_depth', 2, 16)
            model_param['l2_leaf_reg'] = trial.suggest_loguniform('cb_l2_leaf_reg', 1e-8, .1)
            model_param['learning_rate'] = trial.suggest_int('cb_learning_rate', 1, 100)/1000
            model_param['iterations'] = trial.suggest_int('cb_iterations', 1, 10)*100

        ################################# Other ########################################
        return(model_param)

    def _is_model_start_opt_params(self,):
        return(True)


    def get_model_start_opt_params(self,):
        dafault_params = {
            "cb_depth": 6,
            "cb_min_child_samples": 1,
            "cb_learning_rate": 0.03,
            }
        return(dafault_params)

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 cat_features (list)

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
        cat_features (list)
    Return:
        self
    """
    y_train = self.y_format(y_train)

    if cat_features is not None:
        cat_dims = [X_train.columns.get_loc(i) for i in cat_features[:]]
        train_pool = Pool(X_train, label=y_train, cat_features=cat_dims)
    else:
        train_pool = Pool(X_train, label=y_train,)

    params = self.model_param.copy()
    self.model = self._init_model(model_param=params)
    self.model.fit(train_pool, verbose=False, plot=False,)
    train_pool=None
    return self
def get_feature_importance(self, X)

Return

list feature_importance

Expand source code 
def get_feature_importance(self, X,):
    """
    Return:
        list feature_importance
    """
    if not self._is_possible_feature_importance(): 
        raise Exception("Model cannot get feature_importance")

    fe_lst = self.model.get_feature_importance()
    return (pd.DataFrame(fe_lst, index=X.columns, columns=['value']))
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['min_child_samples'] = trial.suggest_int('cb_min_child_samples', 1, 100)
        model_param['depth'] = trial.suggest_int('cb_depth', 4, 10)
        

    ################################# LVL 2 ########################################
    if opt_lvl >= 2:
        model_param['bagging_temperature'] = trial.suggest_int('cb_bagging_temperature', 0, 10,)
        model_param['subsample'] = trial.suggest_discrete_uniform('cb_subsample', 0.1, 1.0, 0.1)
    
    ################################# LVL 3 ########################################
    if opt_lvl >= 3:
        if self._type_of_estimator == 'classifier':
            model_param['objective'] = trial.suggest_categorical('cb_objective', 
                    [
                    'Logloss', 
                    'CrossEntropy',
                    ])

        elif self._type_of_estimator == 'regression':
            model_param['objective'] = trial.suggest_categorical('cb_objective', 
                [
                'MAE',
                'MAPE',
                'Quantile',
                'RMSE',
                ])
    

    ################################# LVL 4 ########################################
    if opt_lvl >= 4:
        model_param['depth'] = trial.suggest_int('cb_depth', 2, 16)
        model_param['l2_leaf_reg'] = trial.suggest_loguniform('cb_l2_leaf_reg', 1e-8, .1)
        model_param['learning_rate'] = trial.suggest_int('cb_learning_rate', 1, 100)/1000
        model_param['iterations'] = trial.suggest_int('cb_iterations', 1, 10)*100

    ################################# Other ########################################
    return(model_param)
def get_model_start_opt_params(self)
Expand source code 
def get_model_start_opt_params(self,):
    dafault_params = {
        "cb_depth": 6,
        "cb_min_child_samples": 1,
        "cb_learning_rate": 0.03,
        }
    return(dafault_params)
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=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_proba(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 not self.is_possible_predict_proba(): 
        raise Exception("Model cannot predict probability distribution")
    return self.model.predict_proba(X)[:, 1]
class CatBoostClassifier (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 CatBoostClassifier(CatBoost):
    _type_of_estimator='classifier'
    __name__ = 'CatBoostClassifier'

Ancestors

Inherited members

class CatBoostRegressor (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 CatBoostRegressor(CatBoost):
    _type_of_estimator='regression'
    __name__ = 'CatBoostRegressor'

Ancestors

Inherited members