[ML] 케라스 딥러닝 예재
Author : tmlab / Date : 2017. 12. 29. 21:21 / Category : Analytics
In [1]:
import pandas as pd
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
import keras
In [2]:
df = pd.read_csv('C:/Users/kimsu/SlowML/creditcard.csv')
df.head(1)
Out[2]:
In [6]:
len(df)
Out[6]:
In [3]:
df['Class'].unique() # 0 = no fraud, 1 = fraudulent
Out[3]:
In [7]:
X = df.iloc[:, :-1].values
y = df.iloc[:, -1].values
In [8]:
X_train, X_test, Y_train, Y_test = train_test_split(X, y, test_size=0.1, random_state=1)
In [9]:
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)
In [10]:
from keras.models import Sequential
from keras.layers import Dense, Dropout
In [11]:
# 각 레이어의 가중치와 바이어스를 초기화하는 옵션이 있음
# keras.initializers.RandomNormal(mean=0.0, stddev=0.05, seed=None)
# keras.initializers.RandomUniform(minval=-0.05, maxval=0.05, seed=None)
clf = Sequential([
Dense(units=16, kernel_initializer='uniform', input_dim=30, activation='relu'),
Dense(units=18, kernel_initializer='uniform', activation='relu'),
Dropout(0.25),
Dense(20, kernel_initializer='uniform', activation='relu'),
Dense(24, kernel_initializer='uniform', activation='relu'),
Dense(1, kernel_initializer='uniform', activation='sigmoid')
])
In [12]:
clf.summary()
In [13]:
clf.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
In [14]:
clf.fit(X_train, Y_train, batch_size=15, epochs=2)
Out[14]:
In [15]:
score = clf.evaluate(X_test, Y_test, verbose=0)
print(clf.metrics_names)
print(score)
In [21]:
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten, Conv2D, pooling
from keras.layers.convolutional import Convolution2D, MaxPooling2D
from keras.utils import np_utils
In [24]:
(X_train, y_train),(X_test, y_test) = mnist.load_data() #다운로드
len(X_train)
Out[24]:
In [26]:
# 인풋을 (데이터개수,28 ,28,1)이 되도록 리쉐이프
X_train = X_train.reshape(X_train.shape[0], 28, 28,1)
X_test = X_test.reshape(X_test.shape[0], 28, 28,1)
# 데이터 타입 변환 및 노멀라이즈
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train /= 255
X_test /= 255
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
print(X_test.shape[0], 'test samples')
In [31]:
# 클래스(라벨)을 범주형 데이터로 변환
print(y_train.shape)
print(y_train[:10])
n_class = 10
In [32]:
Y_train = np_utils.to_categorical(y_train, n_class)
Y_test = np_utils.to_categorical(y_test, n_class)
print(Y_train[:10])
In [107]:
model = Sequential() #CNN 이나 RNN은 시퀀셜
#컨볼루션 레이어
model.add(Convolution2D(32, 3, 3, activation='relu',input_shape=(28,28,1)))
model.add(Convolution2D(32, 3, 3, activation='relu'))
model.add(pooling.MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))
#풀리커넥티트 레이어
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(10, activation='softmax'))
In [109]:
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
model.fit(X_train, Y_train, batch_size=32, nb_epoch=1, verbose=1)
Out[109]:
In [110]:
score = model.evaluate(X_test, Y_test, verbose=0)
print(model.metrics_names)
print(score)
In [1]:
import numpy as np
import matplotlib.pyplot as plt
s = np.sin(2 * np.pi * 0.125 * np.arange(20))
plt.plot(s, 'ro-')
plt.xlim(-0.5, 20.5)
plt.ylim(-1.1, 1.1)
plt.show()
In [2]:
from scipy.linalg import toeplitz
S = np.fliplr(toeplitz(np.r_[s[-1], np.zeros(s.shape[0] - 2)], s[::-1]))
S[:5, :3]
Out[2]:
In [3]:
X_train = S[:-1, :3][:, :, np.newaxis]
Y_train = S[:-1, 3]
X_train.shape, Y_train.shape
Out[3]:
In [5]:
X_train
Out[5]:
In [6]:
Y_train
Out[6]:
In [7]:
plt.subplot(211)
plt.plot([0, 1, 2], X_train[0].flatten(), 'bo-', label="input sequence")
plt.plot([3], Y_train[0], 'ro', label="target")
plt.xlim(-0.5, 4.5)
plt.ylim(-1.1, 1.1)
plt.legend()
plt.title("First sample sequence")
plt.subplot(212)
plt.plot([1, 2, 3], X_train[1].flatten(), 'bo-', label="input sequence")
plt.plot([4], Y_train[1], 'ro', label="target")
plt.xlim(-0.5, 4.5)
plt.ylim(-1.1, 1.1)
plt.legend()
plt.title("Second sample sequence")
plt.tight_layout()
plt.show()
In [37]:
import numpy as np
from keras.models import Sequential
from keras.layers import SimpleRNN, Dense
np.random.seed(0)
model = Sequential()
model.add(SimpleRNN(10, input_dim=1, input_length=3)) # 3개의 값이 각각 1개의 뉴런에 1개씩 들어가고 총 10개 데이터임
model.add(Dense(1))
model.compile(loss='mse', optimizer='sgd')
In [39]:
plt.plot(Y_train, 'ro-', label="target")
plt.plot(model.predict(X_train[:,:,:]), 'bs-', label="output")
plt.xlim(-0.5, 20.5)
plt.ylim(-1.1, 1.1)
plt.legend()
plt.title("Before training")
plt.show()
In [40]:
history = model.fit(X_train, Y_train, nb_epoch=100, verbose=0)
In [41]:
plt.plot(history.history["loss"])
plt.title("Loss")
plt.show()
In [42]:
plt.plot(Y_train, 'ro-', label="target")
plt.plot(model.predict(X_train[:,:,:]), 'bs-', label="output")
plt.xlim(-0.5, 20.5)
plt.ylim(-1.1, 1.1)
plt.legend()
plt.title("After training")
plt.show()
