Unsupervised Learning

# 비지도 학습: 클러스터링과 차원 축소 예제
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs, load_iris
from sklearn.cluster import KMeans, DBSCAN
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
from sklearn.metrics import silhouette_score
from sklearn.manifold import TSNE

# ========================
# 1. K-Means 클러스터링
# ========================
print("=== K-Means Clustering ===")

# 샘플 데이터 생성 (레이블 없이 사용)
X, _ = make_blobs(n_samples=300, centers=4, cluster_std=0.60, random_state=42)

# 스케일링
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

# 최적의 k 찾기 (Elbow Method)
inertias = []
silhouettes = []
k_range = range(2, 10)

for k in k_range:
    kmeans = KMeans(n_clusters=k, random_state=42, n_init=10)
    kmeans.fit(X_scaled)
    inertias.append(kmeans.inertia_)
    silhouettes.append(silhouette_score(X_scaled, kmeans.labels_))

# 최적의 k 선택 (Silhouette 기준)
optimal_k = k_range[np.argmax(silhouettes)]
print(f"Optimal k (by silhouette): {optimal_k}")
print(f"Best silhouette score: {max(silhouettes):.3f}")

# 최적 k로 클러스터링 수행
kmeans_final = KMeans(n_clusters=optimal_k, random_state=42, n_init=10)
clusters = kmeans_final.fit_predict(X_scaled)

print(f"Cluster sizes: {np.bincount(clusters)}")


# ========================
# 2. DBSCAN 클러스터링
# ========================
print("\n=== DBSCAN Clustering ===")

# DBSCAN은 클러스터 수를 자동으로 결정
dbscan = DBSCAN(eps=0.5, min_samples=5)
dbscan_clusters = dbscan.fit_predict(X_scaled)

n_clusters_dbscan = len(set(dbscan_clusters)) - (1 if -1 in dbscan_clusters else 0)
n_noise = list(dbscan_clusters).count(-1)

print(f"DBSCAN clusters found: {n_clusters_dbscan}")
print(f"Noise points: {n_noise}")


# ========================
# 3. PCA 차원 축소
# ========================
print("\n=== PCA Dimensionality Reduction ===")

# Iris 데이터 (4D -> 2D)
iris = load_iris()
X_iris = iris.data  # 레이블 없이 특성만 사용

# PCA 적용
pca = PCA(n_components=2)
X_pca = pca.fit_transform(StandardScaler().fit_transform(X_iris))

print(f"Original dimensions: {X_iris.shape[1]}")
print(f"Reduced dimensions: {X_pca.shape[1]}")
print(f"Explained variance ratio: {pca.explained_variance_ratio_}")
print(f"Total variance explained: {sum(pca.explained_variance_ratio_):.3f}")


# ========================
# 4. 클러스터링 + 시각화
# ========================
class UnsupervisedAnalyzer:
    """비지도 학습 분석 클래스"""

    def __init__(self, random_state=42):
        self.random_state = random_state
        self.scaler = StandardScaler()

    def cluster_analysis(self, X, max_k=10):
        """최적 클러스터 수 분석"""
        X_scaled = self.scaler.fit_transform(X)

        results = {
            'k': [],
            'inertia': [],
            'silhouette': []
        }

        for k in range(2, max_k + 1):
            kmeans = KMeans(n_clusters=k, random_state=self.random_state, n_init=10)
            labels = kmeans.fit_predict(X_scaled)

            results['k'].append(k)
            results['inertia'].append(kmeans.inertia_)
            results['silhouette'].append(silhouette_score(X_scaled, labels))

        return results

    def reduce_and_cluster(self, X, n_clusters=3, reduction='pca'):
        """차원 축소 후 클러스터링"""
        X_scaled = self.scaler.fit_transform(X)

        # 차원 축소
        if reduction == 'pca':
            reducer = PCA(n_components=2, random_state=self.random_state)
        else:  # t-SNE
            reducer = TSNE(n_components=2, random_state=self.random_state,
                          perplexity=30, n_iter=300)

        X_reduced = reducer.fit_transform(X_scaled)

        # 클러스터링
        kmeans = KMeans(n_clusters=n_clusters, random_state=self.random_state, n_init=10)
        labels = kmeans.fit_predict(X_scaled)

        return X_reduced, labels

    def anomaly_detection(self, X, contamination=0.1):
        """이상치 탐지 (Isolation Forest)"""
        from sklearn.ensemble import IsolationForest

        X_scaled = self.scaler.fit_transform(X)

        iso_forest = IsolationForest(
            contamination=contamination,
            random_state=self.random_state
        )

        # -1: anomaly, 1: normal
        predictions = iso_forest.fit_predict(X_scaled)
        anomalies = X[predictions == -1]

        return predictions, anomalies


# 사용 예제
analyzer = UnsupervisedAnalyzer()

# 클러스터 분석
print("\n=== Cluster Analysis ===")
results = analyzer.cluster_analysis(X)
best_k = results['k'][np.argmax(results['silhouette'])]
print(f"Best k: {best_k}, Silhouette: {max(results['silhouette']):.3f}")

# 차원 축소 + 클러스터링
X_2d, labels = analyzer.reduce_and_cluster(X_iris, n_clusters=3)
print(f"\nIris data clustered into {len(set(labels))} groups")
print(f"2D representation shape: {X_2d.shape}")

# 이상치 탐지
predictions, anomalies = analyzer.anomaly_detection(X)
print(f"\nDetected {len(anomalies)} anomalies out of {len(X)} samples")