딥러닝

[calmnis]

# 라이브러리

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

# 데이터 불러오기

df = pd.read_csv('cust_data.csv')

df.info()

df.head()

# 정답지 칼럼 'termination' 분포도
df['termination'].value_counts().plot(kind='bar')

# 문자 타입 칼럼 가져오기

cols = list(df.select_dtypes('object'))

cols 

# 문자 타입 칼럼에 대해 One-Hot-Encoding 수행
df1 = pd.get_dummies(data=df, columns=cols, drop_first=True)

df1.info()

# 문제지, 정답지 분리

X = df1.drop('termination_Y', axis=1).values

y = df1['termination_Y'].values

X.shape, y.shape

# 훈련데이터, 검증데이터 분리

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=42)

X_train.shape, y_train.shape

# 0 ~ 1 값으로 정규화

from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()

X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

X_train[:2], X_test [:2]

X_train.shape, y_train.shape

((5469, 18), (5469,))

------------------------- 딥러닝 모델별 학습, 성능시각화 -------------------------

A. 회귀 DNN layer

• 18개 input layer
• unit 4개 hidden layer
• unit 3개 hidden layer
• 1개 output layser : 이진분류

# 라이브러리

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Activation, Dropout

# 18개 input layer
# unit 4개 hidden layer
# unit 3개 hidden layer 
# 1개 output layser : 회귀

# 모델생성 방법1
model = Sequential()
model.add(Dense(4, activation='relu', input_shape=(18,)))
model.add(Dense(3, activation='relu'))
model.add(Dense(1))

model.summary()

model.compile(optimizer='adam', loss='mse') 

# 모델생성 방법2 - Dropout 과적합 방지

model = Sequential()
model.add(Dense(4, activation='relu', input_shape=(18,)))
model.add(Dropout(0.3))
model.add(Dense(3, activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(1, activation='sigmoid'))

model.summary()

# 컴파일

model.compile(optimizer='adam', loss='mse', metrics=['accuracy']) 

# 학습

history = model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=20, batch_size=16)

# 학습 및 검증 손실(MSE) 시각화
plt.plot(history.history['loss'], label='Train MSE')
plt.plot(history.history['val_loss'], label='Validation MSE')
plt.xlabel('Epoch')
plt.ylabel('Mean Squared Error')
plt.legend()
plt.show()

# 검증 데이터로 예측 수행
y_pred_train = model.predict(X_train)
y_pred_val = model.predict(X_val)

# 학습 MSE 계산
mean_squared_error(y_train, y_pred_train)

# 검증 MSE 계산
mean_squared_error(y_val, y_pred_val)

B. 이진 분류용 DNN layer

• 18개 input layer
• unit 4개 hidden layer
• unit 3개 hidden layer
• 1개 output layser : 이진분류

# 라이브러리

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Activation, Dropout

# 18개 input layer
# unit 4개 hidden layer
# unit 3개 hidden layer 
# 1개 output layser : 이진분류

# 모델생성 방법1
model = Sequential()
model.add(Dense(4, activation='relu', input_shape=(18,)))
model.add(Dense(3, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

model.summary()

model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy']) 

# 모델생성 방법2 - Dropout 과적합 방지

model = Sequential()
model.add(Dense(4, activation='relu', input_shape=(18,)))
model.add(Dropout(0.3))
model.add(Dense(3, activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(1, activation='sigmoid'))

model.summary()

# 컴파일

model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy']) 

# 학습

history = model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=20, batch_size=16)

# 성능시각화

losses = pd.DataFrame(model.history.history)

losses.head()

losses[['loss','val_loss']].plot()

losses[['loss','val_loss', 'accuracy','val_accuracy']].plot()

plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Acc')
plt.legend(['acc', 'val_acc'])
plt.show()

C. 다중 분류용 DNN layer

• 18개 input layer
• unit 5개 hidden layer
• dropout
• unit 4개 hidden layer
• dropout
• 2개 output layser : 다중분류

# 모델생성

model = Sequential()
model.add(Dense(5, activation='relu', input_shape=(18,)))
model.add(Dropout(0.3))
model.add(Dense(4, activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(2, activation='softmax'))

model.summary()

# 컴파일

model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) 

history = model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=20, batch_size=16)

# 성능시각화

losses = pd.DataFrame(model.history.history)

losses.head()

losses[['loss','val_loss']].plot()

losses[['loss','val_loss', 'accuracy','val_accuracy']].plot()

plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Acc')
plt.legend(['acc', 'val_acc'])
plt.show()