无监督学习：聚类/降维/异常检测

大鱼>2026-07-05 10:18

无监督学习：聚类/降维/异常检测

1. 聚类算法

python 复制代码

from sklearn.cluster import KMeans, DBSCAN, AgglomerativeClustering

# K-Means
kmeans = KMeans(n_clusters=3, random_state=42)
labels = kmeans.fit_predict(X)

# 肘部法则选择 K
inertias = []
for k in range(2, 11):
    km = KMeans(n_clusters=k, random_state=42)
    km.fit(X)
    inertias.append(km.inertia_)

# DBSCAN（密度聚类）
dbscan = DBSCAN(eps=0.5, min_samples=5)
labels = dbscan.fit_predict(X)

# 层次聚类
hc = AgglomerativeClustering(n_clusters=3)
labels = hc.fit_predict(X)

2. 降维算法

python 复制代码

from sklearn.decomposition import PCA
from sklearn.manifold import TSNE

# PCA
pca = PCA(n_components=2)
X_pca = pca.fit_transform(X)
print(f"解释方差比: {pca.explained_variance_ratio_}")

# t-SNE（可视化用）
tsne = TSNE(n_components=2, random_state=42, perplexity=30)
X_tsne = tsne.fit_transform(X)

3. 异常检测

python 复制代码

from sklearn.ensemble import IsolationForest
from sklearn.svm import OneClassSVM

# 孤立森林
iso = IsolationForest(contamination=0.1, random_state=42)
outliers = iso.fit_predict(X)  # -1 为异常

# One-Class SVM
ocsvm = OneClassSVM(kernel='rbf', nu=0.1)
outliers = ocsvm.fit_predict(X)

总结

任务	算法	适用场景
聚类	K-Means/DBSCAN	客户分群/图像分割
降维	PCA/t-SNE	可视化/去噪
异常检测	Isolation Forest	欺诈检测/故障诊断