24.05.30 SLP/ MLP

Single-layer Perceptron (SLP) 코딩 실습

• Import
  
  import numpy as np

  import pandas as pd

  import seaborn as sns

  
  

  import matplotlib as mpl

  import matplotlib.pyplot as plt

  import matplotlib_inline.backend_inline

  
  

  import sklearn

  from sklearn.model_selection import train_test_split

  from sklearn.preprocessing import MinMaxScaler

  from sklearn.ensemble import RandomForestClassifier

  from sklearn import metrics

  from sklearn.datasets import load_breast_cancer

  from sklearn.datasets import load_digits

  
  

  import tensorflow

  from tensorflow.keras.utils import to_categorical

  from tensorflow.keras.models import Sequential, load_model

  from tensorflow.keras.layers import Dense

  from tensorflow.keras.optimizers import Adam

  from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping

  from tensorflow.keras.datasets import mnist

  
  

  import random
• Data 불러오기
  
  digits = load_digits()

  digits
• Data info. 확인하기
  
  digits.keys()
• x, y 설정
  
  x_data = digits["images"]

  y_data = digits["target"]
• train/ test data split 
  
  x_train, x_test, y_train, y_test = train_test_split(x_data,

                                                      y_data,

                                                      test_size = 0.20,

                                                      random_state = 42,

                                                      stratify = y_data)
• Scaling 하기
  
  x_train, x_test = x_train / x_train.max(), x_test / x_train.max()
• Data의 shape 확인하기

 

print("x_train: ", x_train.shape)

print("y_train: ", y_train.shape)

print("x_test: ", x_test.shape)

print("y_test: ", y_test.shape)

x_train:  (1437, 8, 8)
y_train:  (1437,)
x_test:  (360, 8, 8)
y_test:  (360,)

# 데이터의 shape을 확인 후 2차원으로 만들어주기 (딥러닝)

x_train_re = np.reshape(x_train, (x_train.shape[0], 64))

x_test_re = np.reshape(x_test, (x_test.shape[0], 64))

x_train_re.shape

(1437, 64)​

• one-hot encoding 하기
  
  y_train = to_categorical(y_train)

  y_train
• modeling하기 
  
  random.seed(209)

  # 입력노드 수 = feature 수

  input_node = 8*8

  # 출력노드 수

  output_node = 10

  
  

  #모델

  model_digit = Sequential()

  model_digit.add(Dense(output_node,

                        activation = "softmax",

                        input_shape = (input_node,)

                       ))
• compliling 
  
  model_digit.compile(loss='categorical_crossentropy',

                       optimizer='adam',

                       metrics = ['accuracy'])
• Hyperparameter 정하기
  
  model_digit_hist = model_digit.fit(x_train_re,

                                     y_train,

                                     epochs=100,

                                     batch_size=10,

                                     validation_split=0.2)
• 시각화해서 확인하기
  
  fig, (ax_loss, ax_acc) = plt.subplots(1, 2, figsize = (8,3))

  
  

  ax_loss.plot(model_digit_hist.history["loss"], label="train_loss")

  ax_loss.plot(model_digit_hist.history["val_loss"], label="val_loss")

  ax_loss.set(xlabel = "epoch",

         ylabel = "loss")

  ax_loss.legend()

  
  

  ax_acc.plot(model_digit_hist.history["accuracy"], label="train_acc")

  ax_acc.plot(model_digit_hist.history["val_accuracy"], label="val_acc")

  ax_acc.set(xlabel = "epoch",

         ylabel = "accuracy")

  ax_acc.legend();
• Accuracy 확인
  
  metrics.accuracy_score(pred_digit, y_test)

 Multi-layer Perceptron (MLP) 코딩 실습

• Data 불러오기
  
  cancer = load_breast_cancer()
• x, y값 정하기
  
  x_data = cancer["data"]

  y_data = cancer["target"]
  - 딥러닝 모델을 만들때는 dataframe으로 fit할 수 없음
• Train/test data split
  
  x_train, x_test, y_train, y_test = train_test_split(x_data,

                                                      y_data,

                                                      test_size = 0.20,

                                                      random_state = 42,

                                                      stratify = y_data)
• Scaling
  
  minmax = MinMaxScaler()

  x_train_scaled = minmax.fit_transform(x_train)

  x_test_scaled = minmax.transform(x_test)
• Data shape 확인
  
  print("x_train: ", x_train.shape)

  print("y_train: ", y_train.shape)

  print("x_test: ", x_test.shape)

  print("y_test: ", y_test.shape)
• Modeling
  
  random.seed(42)

  # 입력노드 수 = feature 수

  input_node = 30

  # 출력노드 수

  output_node = 1

  # 은닉층 노드 수

  first_hidden_node = 10

  
  

  #모델

  model_cancer = Sequential(name = "cancer_classifier")

  model_cancer.add(Dense(first_hidden_node,

                         # kernel_initializer='he_normal',

                         activation = "relu",

                         input_shape=(input_node,)

                         ))

  model_cancer.add(Dense(output_node,

                         # kernel_initializer='he_normal',

                         activation = "sigmoid"

                        ))
• Compling
  
  adam = Adam(learning_rate=0.01)

  
  

  model_cancer.compile(loss='binary_crossentropy',

                       optimizer=adam,

                       metrics = ['accuracy'])
• Hyperparameter 입력
  
  model_cancer_hist = model_cancer.fit(x_train_scaled,

                                       y_train,

                                       epochs=10,

                                       batch_size=10,

                                       validation_split=0.2,

                                       # verbose = 1

                                      )
• Learning 과정 시각화 
  
  fig, (ax_loss, ax_acc) = plt.subplots(1, 2, figsize = (8,3))

  
  

  ax_loss.plot(model_cancer_hist.history["loss"], label="train_loss")

  ax_loss.plot(model_cancer_hist.history["val_loss"], label="val_loss")

  ax_loss.set(xlabel = "epoch", ylabel = "loss")

  ax_loss.legend()

  
  

  ax_acc.plot(model_cancer_hist.history["accuracy"], label="train_acc")

  ax_acc.plot(model_cancer_hist.history["val_accuracy"], label="val_acc")

  ax_acc.set(xlabel = "epoch", ylabel = "accuracy")

  ax_acc.legend();
• Model 평가
  
  pred_cancer = model_cancer.predict(x_test_scaled).flatten() >= 0.5
  #0.5 기준은 바뀔 수 있음
• Accuracy 확인
  
  metrics.accuracy_score(pred_cancer, y_test)