행동 분석으로 고객 유도하기

실전문제로 배우는 머신러닝 - 3주차

• 무료 유저를 유료 멤버로 전환

• 유료 버전의 앱을 등록할 가능성이 없는 유저를 식별하고, 이러한 사용자를 대상으로 마케팅

• 앱을 처음 연 후 처음 24시간 동안의 데이터

• 다운로드 링크 : appdata10.csv

1. 라이브러리 호출

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

2. Data 불러오기

dataset = pd.read_csv('./new_appdata10.csv')

3. Data Pre-Processing

response = dataset["enrolled"]
dataset = dataset.drop(columns="enrolled")

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(dataset, response,
                                                    test_size = 0.2,
                                                    random_state = 0)

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]

train_identity = X_train['user']
X_train = X_train.drop(columns = ['user'])
test_identity = X_test['user']
X_test = X_test.drop(columns = ['user'])

from sklearn.preprocessing import 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

4. Model Building

from sklearn.linear_model import LogisticRegression
classifier = LogisticRegression(random_state = 0, penalty = 'l1')
classifier.fit(X_train, y_train)

y_pred = classifier.predict(X_test)

5. 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 = (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_selection import 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))

# Analyzing Coefficients
pd.concat([pd.DataFrame(dataset.drop(columns = 'user').columns, columns = ["features"]),
           pd.DataFrame(np.transpose(classifier.coef_), columns = ["coef"])
           ],axis = 1)

6. Model Tuning

from sklearn.model_selection import GridSearchCV

penalty = ['l1', 'l2']

C = [0.001, 0.01, 0.1, 1, 10, 100, 1000]

parameters = dict(C=C, penalty=penalty)

grid_search = GridSearchCV(estimator = classifier,
                           param_grid = parameters,
                           scoring = "accuracy",
                           cv = 10,
                           n_jobs = -1)
t0 = time.time()
grid_search = grid_search.fit(X_train, y_train)
t1 = time.time()
print("Took %0.2f seconds" % (t1 - t0))

rf_best_accuracy = grid_search.best_score_
rf_best_parameters = grid_search.best_params_
rf_best_accuracy, rf_best_parameters

7. Grid Search

penalty = ['l1', 'l2']

C = [0.1, 0.5, 0.9, 1, 2, 5]

parameters = dict(C=C, penalty=penalty)

grid_search = GridSearchCV(estimator = classifier,
                           param_grid = parameters,
                           scoring = "accuracy",
                           cv = 10,
                           n_jobs = -1)
t0 = time.time()
grid_search = grid_search.fit(X_train, y_train)
t1 = time.time()
print("Took %0.2f seconds" % (t1 - t0))

rf_best_accuracy = grid_search.best_score_
rf_best_parameters = grid_search.best_params_
rf_best_accuracy, rf_best_parameters
grid_search.best_score_

8. End of Model

final_results = pd.concat([y_test, test_identity], axis = 1).dropna()
final_results['predicted_reach'] = y_pred
final_results = final_results[['user', 'enrolled', 'predicted_reach']].reset_index(drop=True)