当前位置:首页 >> 脚本专栏

python中K-means算法基础知识点

能够学习和掌握编程,最好的学习方式,就是去掌握基本的使用技巧,再多的概念意义,总归都是为了使用服务的,K-means算法又叫K-均值算法,是非监督学习中的聚类算法。主要有三个元素,其中N是元素个数,x表示元素,c(j)表示第j簇的质心,下面就使用方式给大家简单介绍实例使用。

K-Means算法进行聚类分析

km = KMeans(n_clusters = 3)
km.fit(X)
centers = km.cluster_centers_
print(centers)

三个簇的中心点坐标为:

[[5.006 3.428 ]

[6.81276596 3.07446809]

[5.77358491 2.69245283]]

比较一下K-Means聚类结果和实际样本之间的差别:

predicted_labels = km.labels_
fig, axes = plt.subplots(1, 2, figsize=(16,8))
axes[0].scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Set1, 
        edgecolor='k', s=150)
axes[1].scatter(X[:, 0], X[:, 1], c=predicted_labels, cmap=plt.cm.Set1,
        edgecolor='k', s=150)
axes[0].set_xlabel('Sepal length', fontsize=16)
axes[0].set_ylabel('Sepal width', fontsize=16)
axes[1].set_xlabel('Sepal length', fontsize=16)
axes[1].set_ylabel('Sepal width', fontsize=16)
axes[0].tick_params(direction='in', length=10, width=5, colors='k', labelsize=20)
axes[1].tick_params(direction='in', length=10, width=5, colors='k', labelsize=20)
axes[0].set_title('Actual', fontsize=18)
axes[1].set_title('Predicted', fontsize=18)

k-means算法实例扩展内容:

# -*- coding: utf-8 -*- 
"""Excercise 9.4"""
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import sys
import random

data = pd.read_csv(filepath_or_buffer = '../dataset/watermelon4.0.csv', sep = ',')[["密度","含糖率"]].values

########################################## K-means ####################################### 
k = int(sys.argv[1])
#Randomly choose k samples from data as mean vectors
mean_vectors = random.sample(data,k)

def dist(p1,p2):
 return np.sqrt(sum((p1-p2)*(p1-p2)))
while True:
 print mean_vectors
 clusters = map ((lambda x:[x]), mean_vectors) 
 for sample in data:
  distances = map((lambda m: dist(sample,m)), mean_vectors) 
  min_index = distances.index(min(distances))
  clusters[min_index].append(sample)
 new_mean_vectors = []
 for c,v in zip(clusters,mean_vectors):
  new_mean_vector = sum(c)/len(c)
  #If the difference betweenthe new mean vector and the old mean vector is less than 0.0001
  #then do not updata the mean vector
  if all(np.divide((new_mean_vector-v),v) < np.array([0.0001,0.0001]) ):
   new_mean_vectors.append(v) 
  else:
   new_mean_vectors.append(new_mean_vector) 
 if np.array_equal(mean_vectors,new_mean_vectors):
  break
 else:
  mean_vectors = new_mean_vectors 

#Show the clustering result
total_colors = ['r','y','g','b','c','m','k']
colors = random.sample(total_colors,k)
for cluster,color in zip(clusters,colors):
 density = map(lambda arr:arr[0],cluster)
 sugar_content = map(lambda arr:arr[1],cluster)
 plt.scatter(density,sugar_content,c = color)
plt.show()