머신러닝 두 번째 여정: 예측과 분류의 세계로

import numpy as np
import pandas as pd
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score

# 데이터 준비
X = np.random.rand(100, 1) * 10
y = 2 * X + 1 + np.random.randn(100, 1) * 0.5

# 모델 학습
model = LinearRegression()
model.fit(X, y)

# 예측
y_pred = model.predict(X)
print(f"계수: {model.coef_[0][0]:.2f}")
print(f"절편: {model.intercept_[0]:.2f}")

# 다중 변수 데이터
X_multi = np.random.rand(100, 3) * 10
y_multi = 2*X_multi[:, 0] + 1.5*X_multi[:, 1] - 0.5*X_multi[:, 2] + np.random.randn(100) * 0.5

# 다중 선형 회귀
model_multi = LinearRegression()
model_multi.fit(X_multi, y_multi)

# 변수 중요도 확인
for i, coef in enumerate(model_multi.coef_):
    print(f"변수 {i+1} 계수: {coef:.2f}")

from sklearn.linear_model import LogisticRegression
from sklearn.metrics import classification_report, confusion_matrix

# 이진 분류 데이터 생성
X_binary = np.random.randn(200, 2)
y_binary = (X_binary[:, 0] + X_binary[:, 1] > 0).astype(int)

# 로지스틱 회귀 모델
logistic_model = LogisticRegression()
logistic_model.fit(X_binary, y_binary)

# 예측 및 평가
y_pred_binary = logistic_model.predict(X_binary)
print(classification_report(y_binary, y_pred_binary))

from sklearn.datasets import make_classification
from sklearn.multiclass import OneVsRestClassifier

# 다중 클래스 데이터
X_multi_class, y_multi_class = make_classification(
    n_samples=300, n_features=4, n_classes=3, n_clusters_per_class=1
)

# 다중 클래스 로지스틱 회귀
multi_logistic = LogisticRegression(multi_class='ovr')
multi_logistic.fit(X_multi_class, y_multi_class)

# 예측
y_pred_multi = multi_logistic.predict(X_multi_class)
print(classification_report(y_multi_class, y_pred_multi))

from sklearn.model_selection import cross_val_score, KFold

# K-Fold 교차 검증
kfold = KFold(n_splits=5, shuffle=True, random_state=42)
scores = cross_val_score(logistic_model, X_binary, y_binary, cv=kfold)

print(f"교차 검증 점수: {scores.mean():.3f} (+/- {scores.std() * 2:.3f})")

from sklearn.model_selection import GridSearchCV

# 하이퍼파라미터 그리드
param_grid = {
    'C': [0.1, 1, 10, 100],
    'penalty': ['l1', 'l2'],
    'solver': ['liblinear', 'saga']
}

# 그리드 서치
grid_search = GridSearchCV(
    LogisticRegression(), param_grid, cv=5, scoring='accuracy'
)
grid_search.fit(X_binary, y_binary)

print(f"최적 파라미터: {grid_search.best_params_}")
print(f"최고 점수: {grid_search.best_score_:.3f}")

import pandas as pd
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestRegressor

# 주택 가격 데이터 (예시)
data = pd.DataFrame({
    'area': np.random.rand(1000) * 200 + 50,
    'bedrooms': np.random.randint(1, 6, 1000),
    'bathrooms': np.random.randint(1, 4, 1000),
    'age': np.random.randint(0, 50, 1000)
})

# 가격 생성 (실제로는 더 복잡한 관계)
data['price'] = (
    data['area'] * 100 + 
    data['bedrooms'] * 50000 + 
    data['bathrooms'] * 30000 - 
    data['age'] * 2000 + 
    np.random.randn(1000) * 10000
)

# 특성과 타겟 분리
X_house = data[['area', 'bedrooms', 'bathrooms', 'age']]
y_house = data['price']

# 데이터 분할
X_train, X_test, y_train, y_test = train_test_split(
    X_house, y_house, test_size=0.2, random_state=42
)

# 스케일링
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

# 랜덤 포레스트 모델
rf_model = RandomForestRegressor(n_estimators=100, random_state=42)
rf_model.fit(X_train_scaled, y_train)

# 예측 및 평가
y_pred_house = rf_model.predict(X_test_scaled)
mse = mean_squared_error(y_test, y_pred_house)
r2 = r2_score(y_test, y_pred_house)

print(f"평균 제곱 오차: {mse:.2f}")
print(f"R² 점수: {r2:.3f}")