JCardenasRdz

# -*- coding: utf-8 -*-
"""
Spyder Editor

This is a temporary script file.
"""

#%%
from sklearn import metrics
import pandas as pd

JCardenasRdz

# -*- coding: utf-8 -*-

"""
@author: Julio Cardenas-Rodriguez, Ph.D.
          jdatascientist@gmail.com
Objective
- Describe how to :
    1) Use pipelines and feature union in Python
    2) Include cleanining as part of the hyperparameter tunning
    

JCardenasRdz

from sklearn import metrics
# TESTING
print('---'*50)
print( ' The r-squared is :', metrics.regression.r2_score(y_test, model.predict(X_test) ) )

JCardenasRdz

from sklearn.linear_model import LinearRegression
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import Pipeline

# final pipeline
pipeline = Pipeline([("data_cleaner", data_cleaner), ("LR", LinearRegression() )])

# TRAINING
# parameter grid
param_grid = dict(   data_cleaner__cont_cleaner__center = [True,False] ,

JCardenasRdz

#%% 4 create final data cleaner
from sklearn.pipeline import FeatureUnion

#allocate
data_cleaner = FeatureUnion( [  cont_data_cleaner, cat_data_cleaner ])

#train
data_cleaner.fit(X_train)

JCardenasRdz

def create_manufacturer(DataFrame, N_cases = 3):
    """
    return the top_N + 1 manufacturers in one-hot encoding format
    """
    man = DataFrame['name'].apply(lambda x: x.split(' ')[0])
    cases = man.value_counts().index
    
    manufacturer_encoded = man.replace(cases[N_cases::], 
                                       len(cases[N_cases::]) * ['OTHER_MAN'])

JCardenasRdz

#%% 2. Create pipeline to clean contnous data
from sklearn.base import TransformerMixin, BaseEstimator
import pandas as pd

def center_and_scale_data(DataFrame):
    """
    this function does the actual center and scaling
    """
    D = DataFrame.copy()
    for C in D.columns:

JCardenasRdz

# load libraries and data
from sklearn.model_selection import train_test_split as tts
from seaborn import load_dataset
data = load_dataset('mpg').dropna()

# split data
Xdata = data.drop('mpg',axis = 1)
Ydata = data.mpg.values
X_train, X_test, y_train, y_test = tts( Xdata, Ydata, test_size=0.33, random_state=42)

JCardenasRdz

from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split

D= load_boston()

R1, ytest1, yhat1 = train_GradientBoostingRegressor(D.data, D.target,  loss='quantile', alpha = 0.50, n_steps = 50)

R2, ytest2, yhat2 = train_GradientBoostingRegressor(D.data, D.target,  loss='ls' , n_steps = 50)

JCardenasRdz

#modules
from sklearn.metrics.regression import mean_absolute_error as mae
from sklearn.metrics import make_scorer
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.ensemble import GradientBoostingRegressor
from hyperopt import hp, fmin, tpe
from hyperopt.pyll import scope
import numpy as np
  
# hyperopt object for