手写K均值聚类

K-Means-Clustering

• 数据集：链接：https://pan.baidu.com/s/1aEgDpQoeT_kFbwD1YyrCaw
  提取码：2021

• 代码：

  #!/usr/bin/env python
  # coding: utf-8
  
  # In[6]:
  
  
  ## import Module 
  import numpy as np
  import matplotlib.pyplot as plt
  
  
  # In[5]:
  
  
  ## data analysis
  data = np.loadtxt('数据集//testSet.txt')
  data.shape #(80, 2)
  
  
  # In[7]:
  
  
  ## 数据可视化
  plt.scatter(data[:,0],data[:,1])
  plt.show()
  
  
  # In[54]:
  
  
  # 初始化K个质心
  def initCentroids(data, k):
      shape = data[0].shape[0]
      centroids = np.empty((k,shape))
      for i in range(shape):
          centroids[:,i] = (data[:,i].min() + data[:,i].ptp() * np.random.rand(k,1)).flatten() # 转换成一维的
      return centroids
  
  
  # In[118]:
  
  
  initCentroids(data,4)
  
  
  # In[119]:
  
  
  ## 计算质心 和样本距离的函数
  def euclDistance(v1,v2):
      return np.sqrt(np.sum(np.power(v2-v1,2)))
  
  
  # In[147]:
  
  
  ## algorithm core
  def KMeans(data, k):
      row = data.shape[0]
      centroids = initCentroids(data,k)
      #存放 K 个簇
      while(True):
          flag = True
          lists = [[] for i in range(k)]
          for i in range(row):
              ds = 1000000
              cs = 0
              for j in range(k):
                  d = euclDistance(data[i],centroids[j])
                  if d<ds:
                      ds = d
                      cs = j
              #找到 当前样本距离最近的质心，放到 对应集合中
              lists[cs].append(data[i])
          newLists = np.array([np.mean(lists[i], axis = 0) for i in range(k)])
          if np.array_equal(newLists, centroids) == True:
              break
          else:
              centroids = newLists
          #lists 是 划分好的 聚簇，newLists 是 质点集合
      return lists,newLists
  
  
  # In[152]:
  
  
  ## 调用函数
  lists,newLists = KMeans(data,k = 4)
  newLists
  
  
  # In[153]:
  
  
  ## 在原图中画出 最终质点
  plt.scatter(data[:,0],data[:,1])
  plt.scatter(newLists[:,0],newLists[:,1])
  plt.show()