머신 러닝 (20) GMM Clustering

 import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

from sklearn.datasets import load_wine

from sklearn.preprocessing import StandardScaler

from sklearn.decomposition import PCA

from sklearn.mixture import GaussianMixture

# 와인 데이터셋 로드

wine = load_wine()

X = pd.DataFrame(wine.data, columns=wine.feature_names)

# 데이터 정규화

scaler = StandardScaler()

X_scaled = scaler.fit_transform(X)

print(f"데이터 크기: {X.shape}")

print(f"\n데이터 샘플:\n{X.head()}")

 PCA 적용

pca = PCA(n_components=2)

X_pca = pca.fit_transform(X_scaled)

print(f"설명된 분산 비율: {pca.explained_variance_ratio_}")

print(f"총 설명된 분산: {sum(pca.explained_variance_ratio_):.2%}")

# GMM 모델 설정 및 학습

gmm = GaussianMixture(n_components=3, random_state=42)

gmm.fit(X_pca)

# 군집 예측

labels = gmm.predict(X_pca)

print(f"발견된 군집 개수: {len(np.unique(labels))}")

print(f"\n군집 중심점 (평균):\n{gmm.means_}")

# 각 군집별 샘플 개수

for k in range(3):

    count = np.sum(labels == k)

    print(f"클러스터 {k}: {count}개")

# 각 군집에 속할 확률 계산

probs = gmm.predict_proba(X_pca)

print("첫 5개 샘플의 군집별 소속 확률:")

print("Sample | Cluster 0 | Cluster 1 | Cluster 2")

print("-" * 45)

for i in range(5):

    print(f"  {i:2d}   | {probs[i,0]:9.6f} | {probs[i,1]:9.6f} | {probs[i,2]:9.6f}")

plt.figure(figsize=(6, 4))

# 군집 중심점(평균) 추출

centers = gmm.means_

# 군집별 색상 생성

colors = plt.cm.viridis(np.linspace(0, 1, 3))

for k, col in zip(range(3), colors):

    my_members = (labels == k)

    # 군집 데이터 포인트

    plt.plot(X_pca[my_members, 0], X_pca[my_members, 1], 'o',

             markerfacecolor=col, markeredgecolor='k', markersize=8, alpha=0.6)

    # 가우시안 분포의 중심점

    plt.plot(centers[k, 0], centers[k, 1], '*',

             markerfacecolor='red', markeredgecolor='k', markersize=20)

plt.title('GMM Clustering Results (Wine Dataset + PCA)\nComponents: 3')

plt.xlabel('Principal Component 1')

plt.ylabel('Principal Component 2')

plt.grid(True)

plt.show()

fig, axes = plt.subplots(1, 2, figsize=(8, 4))

# GMM 결과

colors_gmm = plt.cm.viridis(np.linspace(0, 1, 3))

for k, col in zip(range(3), colors_gmm):

    my_members = (labels == k)

    axes[0].scatter(X_pca[my_members, 0], X_pca[my_members, 1],

                   c=[col], label=f'Cluster {k}', s=60, edgecolors='k', alpha=0.6)

    axes[0].scatter(centers[k, 0], centers[k, 1],

                   marker='*', c='red', s=300, edgecolors='k')

axes[0].set_title('GMM (3 components)')

axes[0].set_xlabel('PC1')

axes[0].set_ylabel('PC2')

axes[0].legend()

axes[0].grid(True)

# 원본 레이블

colors_true = plt.cm.Set1(np.linspace(0, 1, len(np.unique(wine.target))))

for k, col in zip(np.unique(wine.target), colors_true):

    my_members = (wine.target == k)

    axes[1].scatter(X_pca[my_members, 0], X_pca[my_members, 1],

                   c=[col], label=f'Class {k}', s=60, edgecolors='k', alpha=0.6)

axes[1].set_title(f'True Labels ({len(np.unique(wine.target))} classes)')

axes[1].set_xlabel('PC1')

axes[1].set_ylabel('PC2')

axes[1].legend()

axes[1].grid(True)

plt.tight_layout()

plt.show()