K-means算法通俗原理及Python与R语言的分别实现

   
   
import numpy as np
import random as rd
import matplotlib.pyplot as plt
import math
#数据
dat = np.array([[14,22,15,20,30,18,32,13,23,20,21,22,23,24,35,18],
                [15,28,18,30,35,20,30,15,25,23,24,25,26,27,30,16]])
print(dat)
#聚类中心#
n = len(dat[0])
N = len(dat)n
k = 3
#-------随机产生-----#
center = rd.sample(range(n),k)
center = np.array([dat.T[i] for i in center])
print(‘初始聚类中心为：’)
print(center)
print(‘-----------------------’)
?
#计算聚类中心
def cent(x):
   return(sum(x)/len(x))
?
#计算各点到聚类中心的距离之和
def dist(x):
   #聚类中心
   m0 = cent(x)
   dis = sum(sum((x-m0)2))
   return(dis)
?
#距离
def f(center):
   c0 = []
   c1 = []
   c2 = []
   D = np.arange(k*n).reshape(k,n)
   d0 = center[0]-dat.T
   d1 = center[1]-dat.T
   d2 = center[2]-dat.T
   d = np.array([d0,d1,d2])
   for i in range(k):
      D[i] = sum((d[i]2).T)
   for i in range(n):
      ind = D.T[i].argmin()
      if(ind  0):
         c0.append(i)#分配类别
      else:
         if(ind  1):
            c1.append(i)
         else:
            c2.append(i)
   C0 = np.array([dat.T[i] for i in c0])
   C1 = np.array([dat.T[i] for i in c1])
   C2 = np.array([dat.T[i] for i in c2])
   C = [C0,C1,C2]
   print([c0,c1,c2])
   s = 0
   for i in C:
      s+=dist(i)
   return(s,C)
?
n_max = 50
#初始距离和
print(‘第1次计算！’)
dd,C = f(center)
print(‘距离和为’+str(dd))
print(‘第2次计算！’)
center = [cent(i) for i in C]
Dd,C = f(center)
print(‘距离和为’+str(Dd))
K = 3
?
while(K<n_max):
   #两次差值很小并且计算了一定次数
   if(math.sqrt(dd-Dd)<1 and K>20):
      break;
   print(‘第’+str(K)+‘次计算！’)
   dd = Dd
   print(‘距离和为’+str(dd))
   #当前聚类中心
   center = [cent(i) for i in C]
   Dd,C = f(center)
   K+=1
?
?
#—聚类结果可视化部分—#
?
j = 0
for i in C:
   if(j  0):
      plt.plot(i.T[0],i.T[1],‘ro’)
   if(j  1):
      plt.plot(i.T[0],i.T[1],‘b+’)
   if(j == 2):
      plt.plot(i.T[0],i.T[1],‘g*’)
   j+=1
?
plt.show()

  
  
rm(list = ls())
path <- ‘C:\Users\26015\Desktop\clu.txt’
dat <- read.csv(path,header = FALSE)
dat <- t(dat)
kc <- kmeans(dat,3)
summary(kc)
kc

  
  
K-means clustering with 3 clusters of sizes 8, 3, 5
?
Cluster means:
      [,1]     [,2]
1 21.87500 26.00000
2 32.33333 31.66667
3 15.60000 16.80000