import numpy as np
import scipy.stats as sp
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.preprocessing import minmax_scale, scale, MinMaxScaler, KBinsDiscretizer
from sklearn.compose import ColumnTransformer

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split

from sklearn.linear_model import LogisticRegression, RidgeClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import roc_auc_score, accuracy_score, brier_score_loss

no_cont = 5
no_cat = 5
no_vars = no_cont + no_cat
N= 50000

# Create single dataset to avoid random effects
# Only works for all informative features
X,y = make_classification(n_samples=N, weights=[0.9,0.1], n_clusters_per_class=5,
                              n_features=no_vars, 
                              n_informative=no_vars, 
                              n_redundant=0, n_repeated=0,
                             random_state=123)

binner = KBinsDiscretizer(n_bins=5, encode='ordinal', strategy="quantile", )
X[:,no_cont:] = binner.fit_transform(X[:,no_cont:])

X = pd.DataFrame(X, columns=["X"+str(i) for i in [0,2,4,6,8,1,3,5,7,9]])

X[0:5]

import string
# Efficiently map values to another value with .map(dict)
X.iloc[:,no_cont:] = X.iloc[:,no_cont:].apply(
    lambda x: x.map({i:letter for i,letter in enumerate(string.ascii_uppercase)})
)

X.sort_index(axis=1, inplace=True)

X[0:5]

cat_columns = [1,3,5,7,9]

X.iloc[:,cat_columns] = X.iloc[:,cat_columns].astype("category")

X.dtypes

X[0:5]

X.X9.cat.categories

X.X9.cat.codes[0:5]

X.describe()

X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, 
                                                    test_size=0.5, random_state=123)

from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.impute import SimpleImputer
from sklearn.pipeline import Pipeline

num_columns = [idx for idx in range(X.shape[1]) if idx not in cat_columns]
print(cat_columns, num_columns)
np.all([isinstance(x,int) for x in cat_columns])

num_preproc = Pipeline([
    ('cat_imputer', SimpleImputer(strategy='most_frequent')),
    ("scaler", StandardScaler())
])

ct = ColumnTransformer([
    # (name, transformer, columns)
    # Transformer for categorical variables
    ("onehot", 
         OneHotEncoder(categories='auto', handle_unknown='ignore', ),  
         cat_columns),
    # Transformer for numeric variables
    ("num_preproc",num_preproc, num_columns)
    ], 
    # what to do with variables not specified in the indices?
    remainder="drop")

X_train_transformed = ct.fit_transform(X_train, y)

pd.DataFrame(X_train_transformed)[0:5]

X_train_transformed[:,25:].mean(axis=0)

from sklearn.linear_model import LogisticRegression

pipe = Pipeline([
    ("preprocessing", ct),
    ("classifier", LogisticRegression(C=1, solver='lbfgs'))
])

pipe.fit(X_train, y_train)

pipe.named_steps.classifier.coef_

from sklearn.metrics import roc_auc_score

X_test[0:5]

pred = pipe.predict_proba(X_test)[:,1]

roc_auc_score(y_test, pred)