两个聚类算法

K-means算法：

# -*- coding: utf-8 -*-
# from sklearn.cluster import KMeans
# km = KMeans(n_clusters, init, n_init, max_iter, tol, precompute_distances, verbose, random_state, copy_x, n_jobs, algorithm)
# 本代码采用了 Caliski Harabasz 指标衡量聚类情况
# s(k)= (tr(Bk)*(m-k))/(tr(Wk)*(k-1))
# 其中m为训练集样本数，k为类别数。Bk为类别之间的协方差矩阵，Wk为类别内部数据的协方差矩阵。tr为矩阵的迹。
# 类别内部数据的协方差越小越好，类别之间的协方差越大越好，这样的Calinski-Harabasz分数会高.
# 评估
from sklearn.metrics import calinski_harabasz_score
from sklearn.cluster import KMeans
from sklearn.datasets import make_blobs
import matplotlib.pyplot as plt


# 解决中文显示问题
plt.rcParams['font.sans-serif'] = ['SimSun']  # 指定默认字体
plt.rcParams['axes.unicode_minus'] = False  # 解决保存图像是负号'-'显示为方块的问题

# 生成1000个2维的数据集
# centers = [[-2, -2], [0, 0], [2, 2], [3, 3]]
# centers = [[-3, 2], [0, 0], [1, 7], [5, 2]]
centers = [[-1, -1], [0, 0], [1, 1], [2, 2]]
cluster_std = [0.8, 0.4, 0.4, 0.4]
# cluster_std = [0.1, 0.05, 0.05, 0.05]
# cluster_std = [0.8, 0.4, 0.4, 0.4]
X, y = make_blobs(n_samples=1000, n_features=2, centers=centers,
                  cluster_std=cluster_std, random_state=666)


# 使用K-Means进行聚类，k = {2, 3, 4, 5, 6, 7}
for index, k in enumerate((2, 3, 4, 5, 6, 7)):
    km = KMeans(n_clusters=k, random_state=666)  
    y_pre = km.fit_predict(X)    # 用训练器数据X拟合分类器模型并对训练器数据X进行预测

    plt.subplot(2, 3, index+1)
    # c 为color 不同的聚类用不同的颜色表示其样本。
    # X[:, 0]就是在所有数据中取第0维数据
    plt.scatter(X[:, 0], X[:, 1], c=y_pre)
    plt.xticks([])
    plt.yticks([])
    plt.title('k={},score={:.0f}'.format(
        k, calinski_harabasz_score(X, y_pre)))

plt.show()

谱系聚类算法：

# -*- coding: utf-8 -*-
# 本代码采用了 Caliski Harabasz 指标衡量聚类情况
# s(k)= (tr(Bk)*(m-k))/(tr(Wk)*(k-1))
# 其中m为训练集样本数，k为类别数。Bk为类别之间的协方差矩阵，Wk为类别内部数据的协方差矩阵。tr为矩阵的迹。
# 类别内部数据的协方差越小越好，类别之间的协方差越大越好，这样的Calinski-Harabasz分数会高.

from sklearn.metrics import calinski_harabasz_score
from sklearn.cluster import SpectralClustering
from sklearn.datasets import make_blobs
import matplotlib.pyplot as plt
import time

# 解决中文显示问题
plt.rcParams['font.sans-serif'] = ['SimSun']  # 指定默认字体
plt.rcParams['axes.unicode_minus'] = False  # 解决保存图像是负号'-'显示为方块的问题

# 生成1000个6维的数据集，分为5个簇
X, y = make_blobs(n_samples=1000, n_features=6, centers=5, cluster_std=[
                  1.4, 0.4, 0.3, 0.3, 0.4], random_state=666)

plt.subplot(121)
plt.scatter(X[:, 0], X[:, 1], c='k')
plt.xticks([])
plt.yticks([])
plt.title('样本的原始分布')

# 使用Spectral clustering进行聚类: 使用高斯核对n_cluster和gamma进行调参
start = time.process_time()
best_score, best_k, best_gamma = 0, 0, 0
for gamma in (0.01, 0.1, 1.5):
    for k in (3, 4, 5, 6):
        y_pre = SpectralClustering(n_clusters=k, gamma=gamma).fit_predict(X)
        score = calinski_harabasz_score(X, y_pre)
        print('score={:.3f},k={}, gamma={:.4f}'.format(score, k, gamma))
        if score > best_score:
            best_score = score
            best_gamma = gamma
            best_k = k

print('best_score={:.3f}, best_k={}, best_gamma={:.4f}'.format(
    best_score, best_k, best_gamma))

end = time.process_time()
real_time = end - start
print('spending time: {:.3f}s'.format(real_time))

y_pre = SpectralClustering(n_clusters=best_k, gamma=best_gamma).fit_predict(X)
plt.subplot(122)
plt.scatter(X[:, 0], X[:, 1], c=y_pre)
plt.xticks([])
plt.yticks([])
plt.title('k={},score={:.3f}'.format(best_k, best_score))

plt.show()