분류 모델 (classification model)

1. 이항 분류 모델

from keras.models import Sequential
from keras.layers import Dense, Activation
from keras.optimizers import Adam,RMSprop, SGD

x_data = [[1,2],[2,3],[3,4],[4,3],[3,2],[2,1]]
y_data = [[0],[0],[0],[1],[1],[1]]

# Sequential api
# model = Sequential([
#     Dense(units = 1, input_dim = 2, #input_shape=(2,)
#           Activation('sigmoid'))
# ])

model= Sequential()
model.add(Dense(units = 1, input_dim = 2, activation = 'sigmoid'))
model.compile(optimizer=Adam(learning_rate=0.1), loss='binary_crossentropy', metrics=['accuracy'])
print(model.summary())

model.fit(x=x_data, y=y_data, epochs=100, batch_size=1, verbose=0)
m_eval = model.evaluate(x = x_data, y=y_data, batch_size=1, verbose=1)
print(m_eval)

# 새로운 값으로 결과 얻기
import numpy as np
new_data = [[1,2],[10,7]]
pred = model.predict(new_data, batch_size = 1, verbose =1)
print('예측 결과:', pred)
print('예측 결과:', np.squeeze(np.where(pred > 0.5, 1, 0)))
print('예측 결과:', [1 if i > 0.5 else 0  for i in pred])

# function api ------------
from keras.layers import Input
from keras.models import Model

inputs = Input(shape = (2,), batch_size= 1)
outputs = Dense(units=1, activation='sigmoid')(inputs)
model2 = Model(inputs, outputs)

model2.compile(optimizer=Adam(learning_rate=0.1), loss='binary_crossentropy', metrics=['accuracy'])

model2.fit(x=x_data, y=y_data, epochs=100, batch_size=1, verbose=0)
m_eval2 = model2.evaluate(x = x_data, y=y_data, batch_size=1, verbose=1)
print(m_eval2)

이항 분류 모델은 마지막 Dense가 sigmoid.

loss는 binary_crossentropy를 사용한다.

 

2. 다항 분류 모델

# 다항분류 : 출력값이 softmax 함수를 통해 확률 값으로 여러 개가 출력. 이 중에서 확률값이 가장 큰 인덱스를 분류의 결과로 얻음
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import to_categorical #one-hot encoding을 지원
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf

np.random.seed(1)
# tf.random.set_seed(1)

xdata = np.random.random((1000,12))
ydata = np.random.randint(5, size=(1000,1)) # 범주 5개, 시험과목 0:국어 ~ 4:체육라고 가정
print(xdata[:2])
print(ydata[:2]) # [[2][0]] 학습시 label은 원핫 처리 후 학습에 참여
ydata = to_categorical(ydata, num_classes = 5)
print(ydata[:2])
print(np.argmax(i) for i in ydata[:2])

# 모델
model = Sequential()
model.add(Dense(units = 32, input_shape = (12,), activation='relu'))
model.add(Dense(units = 16, activation='relu'))
model.add(Dense(units = 5, activation='softmax'))

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
print(model.summary())
history= model.fit(xdata, ydata, epochs=2000, batch_size=32, verbose=2, shuffle=True)

model_eval = model.evaluate(xdata, ydata)
print('model_eval :', model_eval)

# 시각화
plt.plot(history.history['loss'], label='loss')
plt.xlabel('epochs')
plt.legend()
plt.show()

plt.plot(history.history['accuracy'], label='accuracy')
plt.xlabel('epochs')
plt.legend()
plt.show()

# 예측
print('예측값:',model.predict(xdata[:5]))
print('예측값:',[np.argmax(i) for i in model.predict(xdata[:5])])
print('예측값:',ydata[:5])
print('예측값:',[np.argmax(i) for i in ydata[:5]])

# 새로운 값으로 예측
x_new = np.random.random([1,12])
print(x_new)
new_pred = model.predict(x_new)
print('분류 결과:', new_pred)
print('분류 결과:', np.sum(new_pred))
print('분류 결과:', np.argmax(new_pred))

CLASSES = np.array(['국어','영어','수학','과학','체육'])
print('분류결과', CLASSES[np.argmax(new_pred)])

다항 분류 모델은 마지막 Dense가 softmax.

loss는 categorical_crossentropy를 사용한다.

 

np.argmax, np.argmin 내용

https://jimmy-ai.tistory.com/72

[Numpy] 파이썬 최대값, 최소값 위치 반환 np.argmax, np.argmin

 

3. 한 가지를 이항, 다항 해보기 (이핳을 다항분류로 처리 할 수 있다.)

# 이항분류는 다항분류로도 처리 가능
# diabets dataset

import numpy as np
from keras.models import Sequential
from keras.layers import Dense

dataset = np.loadtxt('../testdata/diabetes.csv', delimiter=',')
print(dataset.shape)
print(dataset[:1])
print(dataset[:,-1])

# 이항분류 : sigmoid

from sklearn.model_selection import train_test_split
x_train, x_test, y_train, y_test = train_test_split(dataset[:,0:8],dataset[:,-1],
                                                    test_size=0.3,shuffle=True, random_state=123)
print(x_train.shape, x_test.shape, y_train.shape, y_test.shape)

model = Sequential()
model.add(Dense(64, input_dim=8, activation='relu'))
model.add(Dense(32, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['acc'])
model.fit(x_train, y_train, epochs=100, batch_size=32, verbose=0, validation_split=0.2)
scores = model.evaluate(x_test, y_test)
print('이항분류 모델 성능: %s : %.2f%%'%(model.metrics_names[1],scores[1]*100))
print('이항분류 모델 성능: %s : %.2f'%(model.metrics_names[0],scores[0]))
pred = model.predict([[-0.294,0.487,0.180,-0.292,0.,0.00149,-0.5311,-0.0333,]])
print('예측 결과:',pred, ' ',np.where(pred > 0.5, 1, 0))

print('------------------')
# 위의 내용을 다항분류로 처리 
from keras.utils import to_categorical #one-hot encoding을 지원

# label은 원핫 처릴
y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
print('y_train:', y_train[0])

model = Sequential()
model.add(Dense(64, input_dim = 8, activation='relu'))
model.add(Dense(32, activation='relu'))
model.add(Dense(2, activation='softmax'))

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc'])
model.fit(x_train, y_train, epochs=100, batch_size=32, verbose=0, validation_split=0.2)
scores = model.evaluate(x_test, y_test)
print('이항분류 모델 성능: %s : %.2f%%'%(model.metrics_names[1],scores[1]*100))
print('이항분류 모델 성능: %s : %.2f'%(model.metrics_names[0],scores[0]))
pred = model.predict([[-0.294,0.487,0.180,-0.292,0.,0.00149,-0.5311,-0.0333,]])
print('예측 결과:',pred, ' ',np.argmax(pred))

 

원핫 처리 하고 싶지않다면

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

compile에 sparse를 넣어준다.