딥러닝 앙상블

gistfile1.py

import numpy as np
from sklearn.ensemble import VotingClassifier
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.preprocessing import StandardScaler
from skorch import NeuralNetClassifier
import torch
import torch.nn as nn

# 샘플 데이터 생성
X, y = make_classification(n_samples=1000, n_features=20, n_classes=2, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# 데이터 정규화
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

# PyTorch 모델 정의
class Net1(nn.Module):
    def __init__(self):
        super(Net1, self).__init__()
        self.fc1 = nn.Linear(20, 64)
        self.dropout1 = nn.Dropout(0.3)  # Dropout 추가
        self.fc2 = nn.Linear(64, 32)
        self.dropout2 = nn.Dropout(0.2)
        self.fc3 = nn.Linear(32, 2)
    
    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.dropout1(x)
        x = torch.relu(self.fc2(x))
        x = self.dropout2(x)
        x = self.fc3(x)
        return x

class Net2(nn.Module):
    def __init__(self):
        super(Net2, self).__init__()
        self.fc1 = nn.Linear(20, 128)
        self.dropout1 = nn.Dropout(0.4)  # Dropout 추가
        self.fc2 = nn.Linear(128, 64)
        self.dropout2 = nn.Dropout(0.3)
        self.fc3 = nn.Linear(64, 2)
    
    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.dropout1(x)
        x = torch.relu(self.fc2(x))
        x = self.dropout2(x)
        x = self.fc3(x)
        return x

# skorch로 래핑, 하이퍼파라미터 다양화
model_1 = NeuralNetClassifier(
    Net1,
    max_epochs=20,  # 에포크 증가
    lr=0.001,
    optimizer=torch.optim.Adam,  # Adam 최적화
    verbose=0
)
model_2 = NeuralNetClassifier(
    Net2,
    max_epochs=20,
    lr=0.002,  # 다른 학습률
    optimizer=torch.optim.Adam,
    verbose=0
)

# VotingClassifier 생성 (가중치 적용)
ensemble = VotingClassifier(
    estimators=[
        ('model_1', model_1),
        ('model_2', model_2)
    ],
    voting='soft',
    weights=[0.6, 0.4]  # model_1에 더 높은 가중치 부여
)

# 교차 검증으로 성능 평가
cv_scores = cross_val_score(ensemble, X_train.astype(np.float32), y_train, cv=5, scoring='accuracy')
print(f"교차 검증 정확도: {cv_scores.mean():.4f} (±{cv_scores.std():.4f})")

# 모델 학습
ensemble.fit(X_train.astype(np.float32), y_train)

# 테스트 데이터로 최종 평가
print("앙상블 모델 테스트 정확도:", ensemble.score(X_test.astype(np.float32), y_test))