k-means——平面上100个样本点的聚类分析(通俗易懂)

import numpy as np
# 构造数据集
x = np.linspace(0,99,100)
y = np.linspace(100,199,100)

aa = 0      # aa变量是为了记录，迭代次数
k = 2       # 指定将数据分为几个类别
n = len(x)  # 数据集的个数 

# 1、随机选取两个点，作为初始的类中心；
center0 = np.array([x[0],y[0]])
center1 = np.array([x[1],y[1]])

dis = np.zeros([n,k+1])
while aa >= 0:
    # 2、求各样本到各类中心的距离；
    for i in range(n):
        dis[i,0] = np.sqrt((x[i]-center0[0])**2+(y[i]-center0[1])**2)
        dis[i,1] = np.sqrt((x[i]-center1[0])**2+(y[i]-center1[1])**2)
        # 3、归类：将样本归类为，距离其最近的类中的所属类；
        dis[i,2] = np.argmin(dis[i,:2])
    # 4、再次计算各类样本的均值，作为新的类中心；
    index0 = dis[:,2] == 0
    index1 = dis[:,2] == 1
    center0_new = np.array([x[index0].mean(),y[index0].mean()])
    center1_new = np.array([x[index1].mean(),y[index1].mean()])
    # 5、判断类中心，是否发生变化。如果发生变化，就回到第2步；否则，break退出循环；
    if all((center0 == center0_new) & (center1 == center1_new)):
        break
    center0 = center0_new
    center1 = center1_new
    aa += 1
print(len(dis[dis[:,2] == 0]),len(dis[dis[:,2] == 1]))
print(center0,center1,aa)