딥러닝 예제

[calmnis]

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import to_categorical

# 가상의 데이터 생성
data = {
    'feature_1': np.random.rand(1000),
    'feature_2': np.random.rand(1000),
    'label': np.random.choice(['A', 'B', 'C'], 1000)
}
df = pd.DataFrame(data)

# 레이블 인코딩
label_encoder = LabelEncoder()
df['label_encoded'] = label_encoder.fit_transform(df['label'])

# 특성과 레이블 분리
X = df[['feature_1', 'feature_2']].values
y = df['label_encoded'].values

# 레이블을 원-핫 인코딩
y_categorical = to_categorical(y)

# 학습 데이터와 테스트 데이터 분리
X_train, X_test, y_train, y_test = train_test_split(X, y_categorical, test_size=0.2, random_state=42)

# 모델 구성
model = Sequential()
model.add(Dense(64, input_dim=X_train.shape[1], activation='relu'))
model.add(Dense(64, activation='relu'))
model.add(Dense(len(label_encoder.classes_), activation='softmax'))  # 출력 노드 수는 클래스 수와 동일해야 함

# 모델 컴파일
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

# 모델 학습
history = model.fit(X_train, y_train, epochs=20, batch_size=32, validation_split=0.2)

• loss: 학습 loss.
• val_loss: 검증 loss.
• accuracy: 학습 accuracy.
• val_accuracy: 검증 accuracy.

# 시각화 예시

plt.plot(history.history['loss'], label='Train Loss')

plt.plot(history.history['val_loss'], label='Validation Loss')

plt.xlabel('Epochs') # X축 라벨

plt.ylabel('Loss') # Y축 라벨

plt.legend() # 범례 표시 - label값

plt.show()

# 모델 평가
loss, accuracy = model.eval‎uate(X_test, y_test)
print(f'Test loss: {loss}')
print(f'Test accuracy: {accuracy}')