Kaggle 신용카드 사기 검출

Kaggle 신용카드 사기 검출

https://www.kaggle.com/mlg-ulb/creditcardfraud

Credit Card Fraud Detection

Credit Card Fraud Detection

• creditcard.csv (284,807 * 31)
• Class : 0 (정상), 1 (사기)
• 사기 검출(Fraud Detection), 이상 탐지(Anomaly Detection)

---

1. Google Drive Mount

#'creditCardFraud.zip' 파일을 구글드라이브에 업로드 후 진행

from google.colab import drive

drive.mount('/content/drive')

!ls -l '/content/drive/My Drive/Colab Notebooks/datasets/creditCardFraud.zip'

---

2. Data Preprocessing

2-1. Unzip 'creditCardFraud.zip'

!unzip /content/drive/My\ Drive/Colab\ Notebooks/datasets/creditCardFraud.zip

!ls -l

---

2-2. 데이터 읽어오기

import pandas as pd

DF = pd.read_csv('creditcard.csv')

DF.info()

---

DF.head()

#0 (정상) Class와 1 (사기) Class 개수

DF.Class.value_counts()

#0 (정상) Class와 1 (사기) Class 비율

(DF.Class.value_counts() / DF.shape[0]) * 100

---

2-3. Time 열(Column) 삭제

DF.drop('Time', axis = 1, inplace = True)

DF.head(1)

---

2-4. train_test_split( )

X = DF.iloc[:,:-1]
y = DF.iloc[:, -1]

X.shape, y.shape

((284807, 29), (284807,))

from sklearn.model_selection import train_test_split 

X_train, X_test, y_train, y_test = train_test_split(X, y,
                                                    test_size = 0.3,
                                                    random_state = 2045,
                                                    stratify = y)

X_train.shape, y_train.shape, X_test.shape, y_test.shape

((199364, 29), (199364,), (85443, 29), (85443,))

#Train_Data와 Test_Data의 1 (부정) 비율이 균형

print('Train_Data :','\n', (y_train.value_counts() / y_train.shape[0]) * 100)
print('Test_Data :','\n', (y_test.value_counts() / y_test.shape[0]) * 100)

---

3. Keras Modeling

3-1. Model Define

import tensorflow
from tensorflow.keras import models
from tensorflow.keras import layers

ccfd = models.Sequential()
ccfd.add(layers.Dense(128, activation = 'relu', input_shape = (29,)))
ccfd.add(layers.Dense(64, activation = 'relu'))
ccfd.add(layers.Dense(32, activation = 'relu'))
ccfd.add(layers.Dense(1, activation = 'sigmoid'))

#모델 구조 확인

ccfd.summary()

---

3-2. Model Compile

ccfd.compile(loss = 'binary_crossentropy',
             optimizer = 'adam',
             metrics = ['Recall'])

---

3-3. Model Fit

%%time

Hist_ccfd = ccfd.fit(X_train, y_train,
                     epochs = 50,
                     batch_size = 1024,
                     validation_data = (X_test, y_test))

---

3-4.학습 결과 시각화

import matplotlib.pyplot as plt

epochs = range(1, len(Hist_ccfd.history['loss']) + 1)

plt.figure(figsize = (9, 6))
plt.plot(epochs, Hist_ccfd.history['loss'])
plt.plot(epochs, Hist_ccfd.history['val_loss'])
plt.title('Training & Validation Loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.grid()
plt.show()

import matplotlib.pyplot as plt

epochs = range(1, len(Hist_ccfd.history['recall']) + 1)

plt.figure(figsize = (9, 6))
plt.plot(epochs, Hist_ccfd.history['recall'])
plt.plot(epochs, Hist_ccfd.history['val_recall'])
plt.title('Training & Validation Recall')
plt.xlabel('Epochs')
plt.ylabel('Recall')
plt.legend(['Training Recall', 'Validation Recall'])
plt.grid()
plt.show()

---

3-5. Model Evaluate

loss, recall = ccfd.evaluate(X_test, y_test)

print('Loss = {:.5f}'.format(loss))
print('Recall = {:.5f}'.format(recall))

---

3-6. Model Predict

import numpy as np

y_hat = np.round(ccfd.predict(X_test))

y_hat

---

from sklearn.metrics import confusion_matrix

confusion_matrix(y_test, y_hat)

from sklearn.metrics import accuracy_score, precision_score, recall_score

print(accuracy_score(y_test, y_hat))
print(precision_score(y_test, y_hat, pos_label = 1))
print(recall_score(y_test, y_hat, pos_label = 1))