kmeans聚类簇个数选择

借助sklearn库实现kmeans聚类和轮廓系数计算

from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score
K = range(2,20) # 设置主题个数区间
coef = [] 
for k in K:
    km = KMeans(n_clusters=k,random_state = 0).fit(feature) #构建kmeans模型并训练
    score = silhouette_score(feature, km.labels_,sample_size=1000) # 计算对应模型的轮廓系数
    coef.append(score)

可视化：

import matplotlib.pyplot as plt
%matplotlib inline
plt.plot(K,coef) # K为x轴输出，coef是y轴输出
plt.show()