머신 러닝 (21) SVC

 import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

from sklearn.datasets import fetch_openml

from sklearn.model_selection import train_test_split, StratifiedKFold

from sklearn.preprocessing import StandardScaler

from sklearn.feature_selection import RFECV

from sklearn.svm import SVC

from sklearn.pipeline import Pipeline

from sklearn.metrics import accuracy_score, classification_report

print("Loading Pima Indians Diabetes Dataset...")

pima = fetch_openml(name='diabetes', version=1, as_frame=True)

X = pima.data

# 'tested_negative'를 0으로, 'tested_positive'를 1로 매핑하여 정수형으로 변환

y = pima.target.map({'tested_negative': 0, 'tested_positive': 1}).astype(int) #pima.target ->Series

print(f"데이터 크기: {X.shape}")

print(f"피처 목록: {list(X.columns)}")

print(f"\n클래스 분포:\n{y.value_counts()}")

# 데이터 분리 (stratify로 클래스 비율 유지)

X_train, X_test, y_train, y_test = train_test_split(

    X, y, test_size=0.2, random_state=42, stratify=y

)

# 피처 선택용 선형 SVC

svc_linear = SVC(kernel="linear", random_state=42)

# RFECV 설정

rfecv = RFECV(

    estimator=svc_linear,

    step=1,

    cv=StratifiedKFold(5),

    scoring='accuracy',

    min_features_to_select=1

)

# Pipeline 구축

pipe_svc = Pipeline([

    ('scaler', StandardScaler()),

    ('feature_selection', rfecv),

    ('classifier', SVC(kernel='rbf', random_state=42))

])

print("Pipeline 구성 완료")

print("\nPipeline 구조:")

print(pipe_svc)

# 학습

print("Training SVC Pipeline with RFECV...")

print("(교차검증 수행 중... 시간이 소요됩니다)")

pipe_svc.fit(X_train, y_train)

print("\n학습 완료!")

# 선택된 피처 정보 추출

fs_step = pipe_svc.named_steps['feature_selection']

n_features = fs_step.n_features_

selected_features = X.columns[fs_step.support_].tolist()

print("[결과 진단]")

print(f"알고리즘이 선택한 최적의 피처 개수: {n_features}개")

print(f"\n선택된 피처 목록:")

for i, feat in enumerate(selected_features, 1):

    print(f"  {i}. {feat}")

# 피처별 순위 확인

feature_ranking = pd.DataFrame({

    'Feature': X.columns,

    'Rank': fs_step.ranking_,

    'Selected': fs_step.support_

}).sort_values('Rank')

print("\n피처별 순위 (1이 선택됨):")

print(feature_ranking)

# 예측 및 평가

y_pred = pipe_svc.predict(X_test)

accuracy = accuracy_score(y_test, y_pred)

print(f"Test Accuracy: {accuracy:.4f}")

print("\n--- Classification Report ---")

print(classification_report(y_test, y_pred))

print("선형 SVC 생성 완료 (피처 선택용)")

# 선형 커널만 사용하는 Pipeline

pipe_linear_only = Pipeline([

    ('scaler', StandardScaler()),

    ('feature_selection', RFECV(

        estimator=SVC(kernel="linear", random_state=42),

        step=1,

        cv=StratifiedKFold(5),

        scoring='accuracy',

        min_features_to_select=1

    )),

    ('classifier', SVC(kernel='linear', random_state=42))

])

print("선형 커널 전용 Pipeline 학습 중...")

pipe_linear_only.fit(X_train, y_train)

y_pred_linear = pipe_linear_only.predict(X_test)

accuracy_linear = accuracy_score(y_test, y_pred_linear)

# 비교

print("\n커널 전략 비교:")

print(f"Linear 전용: Accuracy = {accuracy_linear:.4f}")

print(f"Linear(선택) + RBF(예측): Accuracy = {accuracy:.4f}")

print(f"\n정확도 향상: {accuracy - accuracy_linear:+.4f}")

# RFECV 성능 변화 시각화

plt.figure(figsize=(10, 6))

plt.xlabel("Number of Features Selected")

plt.ylabel("Cross Validation Score (Accuracy)")

cv_scores = fs_step.cv_results_['mean_test_score']

plt.plot(

    range(1, len(cv_scores) + 1),

    cv_scores,

    marker='o',

    linewidth=2,

    markersize=8

)

# 최적 지점 표시

optimal_score = cv_scores[n_features - 1]

plt.axvline(x=n_features, color='r', linestyle='--',

            label=f'Optimal: {n_features} features')

plt.scatter([n_features], [optimal_score],

            color='red', s=200, zorder=5, marker='*')

plt.title("RFECV (Linear SVC): Accuracy vs Number of Features")

plt.legend()

plt.grid(True, alpha=0.3)

plt.tight_layout()

plt.show()

# 피처 개수별 점수 분석

cv_results_df = pd.DataFrame({

    'Num_Features': range(1, len(cv_scores) + 1),

    'Mean_Score': cv_scores,

    'Std_Score': fs_step.cv_results_['std_test_score']

})

print("\n피처 개수별 교차검증 점수:")

print(cv_results_df)

print(f"\n최고 점수: {cv_results_df['Mean_Score'].max():.4f}")

print(f"최적 선택: {n_features}개 피처 (Score: {optimal_score:.4f})")