재정 습관의 분석을 통해 해지율 최소화 하기

실전문제로 배우는 머신러닝 - 4주차

Import Library

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sn

dataset = pd.read_csv('churn_data.csv') # Users who were 60 days enrolled, churn in the next 30

EDA

dataset.head(5) # Viewing the Data
dataset.columns
dataset.describe() # Distribution of Numerical Variables

dataset[dataset.credit_score < 300]
dataset = dataset[dataset.credit_score >= 300]

dataset.isna().any()
dataset.isna().sum()
dataset = dataset.drop(columns = ['credit_score', 'rewards_earned'])

dataset2 = dataset.drop(columns = ['user', 'churn'])
fig = plt.figure(figsize=(15, 12))
plt.suptitle('Histograms of Numerical Columns', fontsize=20)
for i in range(1, dataset2.shape[1] + 1):
    plt.subplot(6, 5, i)
    f = plt.gca()
    f.axes.get_yaxis().set_visible(False)
    f.set_title(dataset2.columns.values[i - 1])

    vals = np.size(dataset2.iloc[:, i - 1].unique())
    
    plt.hist(dataset2.iloc[:, i - 1], bins=vals, color='#3F5D7D')
plt.tight_layout(rect=[0, 0.03, 1, 0.95])

dataset2 = dataset[['housing', 'is_referred', 'app_downloaded',
                    'web_user', 'app_web_user', 'ios_user',
                    'android_user', 'registered_phones', 'payment_type',
                    'waiting_4_loan', 'cancelled_loan',
                    'received_loan', 'rejected_loan', 'zodiac_sign',
                    'left_for_two_month_plus', 'left_for_one_month', 'is_referred']]
fig = plt.figure(figsize=(15, 12))
plt.suptitle('Pie Chart Distributions', fontsize=20)
for i in range(1, dataset2.shape[1] + 1):
    plt.subplot(6, 3, i)
    f = plt.gca()
    f.axes.get_yaxis().set_visible(False)
    f.set_title(dataset2.columns.values[i - 1])
   
    values = dataset2.iloc[:, i - 1].value_counts(normalize = True).values
    index = dataset2.iloc[:, i - 1].value_counts(normalize = True).index
    plt.pie(values, labels = index, autopct='%1.1f%%')
    plt.axis('equal')
fig.tight_layout(rect=[0, 0.03, 1, 0.95])

Exploring Uneven Features

dataset[dataset2.waiting_4_loan == 1].churn.value_counts()
dataset[dataset2.cancelled_loan == 1].churn.value_counts()
dataset[dataset2.received_loan == 1].churn.value_counts()
dataset[dataset2.rejected_loan == 1].churn.value_counts()
dataset[dataset2.left_for_one_month == 1].churn.value_counts()

Correlation with Response Variable

dataset2.drop(columns = ['housing', 'payment_type',
                         'registered_phones', 'zodiac_sign']
    ).corrwith(dataset.churn).plot.bar(figsize=(20,10),
              title = 'Correlation with Response variable',
              fontsize = 15, rot = 45,
              grid = True)

Correlation Matrix

sn.set(style="white")
corr = dataset.drop(columns = ['user', 'churn']).corr()
mask = np.zeros_like(corr, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True

f, ax = plt.subplots(figsize=(18, 15))

cmap = sn.diverging_palette(220, 10, as_cmap=True)
sn.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5, cbar_kws={"shrink": .5})

Removing Correlated Fields

dataset = dataset.drop(columns = ['app_web_user'])
dataset.to_csv('new_churn_data.csv', index = False)

Import Data

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sn

dataset = pd.read_csv('churn_data.csv') # Users who were 60 days enrolled, churn in the next 30
dataset = dataset.drop(columns = ['app_web_user'])

Data Preparation

user_identifier = dataset['user']
dataset = dataset.drop(columns = ['user'])

# Cleaning Data
dataset[dataset.credit_score < 300]
dataset = dataset[dataset.credit_score >= 300]

# Removing NaN
dataset.isna().any()
dataset.isna().sum()
dataset = dataset.drop(columns = ['credit_score', 'rewards_earned'])

One Hot Encoding

dataset.housing.value_counts()
dataset.groupby('housing')['churn'].nunique().reset_index()
dataset= pd.get_dummies(dataset)
dataset.columns
dataset= dataset.drop(columns= ['housing_na', 'zodiac_sign_na', 'payment_type_na'])

Splitting the dataset into the Training set and Test set

from sklearn.model_selectionimport train_test_split
X_train, X_test, y_train, y_test= train_test_split(dataset.drop(columns= 'churn'), dataset['churn'],
                                                    test_size= 0.2,
                                                    random_state= 0)

Balancing the Training Set

import random

y_train.value_counts()

pos_index= y_train[y_train.values== 1].index
neg_index= y_train[y_train.values== 0].index

if len(pos_index)> len(neg_index):
    higher= pos_index
    lower= neg_index
else:
    higher= neg_index
    lower= pos_index

random.seed(0)
higher= np.random.choice(higher, size=len(lower))
lower= np.asarray(lower)
new_indexes= np.concatenate((lower, higher))

X_train= X_train.loc[new_indexes, ]
y_train= y_train[new_indexes]

Feature Scaling

from sklearn.preprocessingimport StandardScaler
sc_X= StandardScaler()
X_train2= pd.DataFrame(sc_X.fit_transform(X_train))
X_test2= pd.DataFrame(sc_X.transform(X_test))
X_train2.columns= X_train.columns.values
X_test2.columns= X_test.columns.values
X_train2.index= X_train.index.values
X_test2.index= X_test.index.values
X_train= X_train2
X_test= X_test2

Model Building

Fitting Model to the Training Set

from sklearn.linear_modelimport LogisticRegression

classifier= LogisticRegression(random_state= 0)
classifier.fit(X_train, y_train)

y_pred = classifier.predict(X_test)

from sklearn.metrics import confusion_matrix, accuracy_score, f1_score, precision_score, recall_score
cm = confusion_matrix(y_test, y_pred)
accuracy_score(y_test, y_pred)
precision_score(y_test, y_pred) # tp / (tp + fp)
recall_score(y_test, y_pred) # tp / (tp + fn)
f1_score(y_test, y_pred)

df_cm = pd.DataFrame(cm, index = (0, 1), columns = (0, 1))
plt.figure(figsize = (10,7))
sn.set(font_scale=1.4)
sn.heatmap(df_cm, annot=True, fmt='g')
print("Test Data Accuracy: %0.4f" % accuracy_score(y_test, y_pred))

from sklearn.model_selectionimport cross_val_score
accuracies= cross_val_score(estimator= classifier, X= X_train, y= y_train, cv= 10)
print("SVM Accuracy: %0.3f (+/- %0.3f)"% (accuracies.mean(), accuracies.std()* 2))

pd.concat([pd.DataFrame(X_train.columns, columns = ["features"]),
           pd.DataFrame(np.transpose(classifier.coef_), columns = ["coef"])
           ],axis = 1)

Feature Selection

Recursive Feature Elimination

from sklearn.feature_selection import RFE
from sklearn.linear_model import LogisticRegression

rfe = RFE(classifier, step = 20)
rfe = RFE(classifier, step = 20)
rfe = rfe.fit(X_train, y_train)

New Correlation Matrix

sn.set(style="white")
corr = X_train[X_train.columns[rfe.support_]].corr()
mask = np.zeros_like(corr, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True

f, ax = plt.subplots(figsize=(18, 15))

# Generate a custom diverging colormap
cmap = sn.diverging_palette(220, 10, as_cmap=True)

# Draw the heatmap with the mask and correct aspect ratio
sn.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5, cbar_kws={"shrink": .5})

Fitting Model to the Training Set

from sklearn.linear_model import LogisticRegression
classifier = LogisticRegression()
classifier.fit(X_train[X_train.columns[rfe.support_]], y_train)

# Predicting Test Set
y_pred = classifier.predict(X_test[X_train.columns[rfe.support_]])

Evaluating Results

from sklearn.metrics import confusion_matrix, accuracy_score, f1_score, precision_score, recall_score
cm = confusion_matrix(y_test, y_pred)
accuracy_score(y_test, y_pred)
precision_score(y_test, y_pred) # tp / (tp + fp)
recall_score(y_test, y_pred) # tp / (tp + fn)
f1_score(y_test, y_pred)

df_cm = pd.DataFrame(cm, index = (1, 0), columns = (1, 0))
plt.figure(figsize = (10,7))
sn.set(font_scale=1.4)
sn.heatmap(df_cm, annot=True, fmt='g')
print("Test Data Accuracy: %0.4f" % accuracy_score(y_test, y_pred))

Applying k-Fold Cross Validation

from sklearn.model_selectionimport cross_val_score
accuracies= cross_val_score(estimator= classifier,
                             X= X_train[X_train.columns[rfe.support_]],
                             y= y_train, cv= 10)
print("SVM Accuracy: %0.3f (+/- %0.3f)"% (accuracies.mean(), accuracies.std()* 2))

Analyzing Coefficients

pd.concat([pd.DataFrame(X_train[X_train.columns[rfe.support_]].columns, columns = ["features"]),
           pd.DataFrame(np.transpose(classifier.coef_), columns = ["coef"])
           ],axis = 1)

Formatting Final Results

final_results= pd.concat([y_test, user_identifier], axis= 1).dropna()
final_results['predicted_churn']= y_pred
final_results= final_results[['user', 'churn', 'predicted_churn']].reset_index(drop=True)