【Python】k-means算法实现

# -*- coding: utf-8 -*-
import  math
import  random
import matplotlib.pyplot as plt
from matplotlib import colors as m_colors
#生成样本点
def gen_random_sample(n_feat,lower,upper):
sample=[random.uniform(lower,upper) for _ in range(n_feat)]
return sample
#样本个数
n_samples=1000
#特征个数（特征维度）
n_feat=2
#特征数值范围
lower=0
upper=200
#聚类个数
n_cluster=3
#聚类样本点
samples=[gen_random_sample(n_feat,lower,upper) for _ in range(n_samples)]

#收敛阈值
cutoff=0.1
#聚类中心
cluster=[samples[random.randint(0,n_samples)] for _ in range(n_cluster)]

#重新计算聚类中心：
new_cluster=[]
#计算两点间距离
def cal_distance(a,b):
s_sum=0
for i in range(n_feat):
s_sum+=pow(a[i]-b[i],2)
dis=math.sqrt(s_sum)
return dis
cutoff_dis=[1 for _ in cluster]
print(max(cutoff_dis))
#迭代次数
n_loop=1
while(max(cutoff_dis)> cutoff ):

# 样本点图
#for i_sample in samples:
#x_sample.append(i_sample[0])
# y_sample.append(i_sample[1])
#plt.scatter(x_sample, y_sample)
# 画出聚类中心
#for cluster_point in cluster:
#plt.scatter(cluster_point[0], cluster_point[1], c='red')
#plt.title('samples&clusters(iteration:{})'.format(n_loop))
#plt.show()
#分类lists
lists=[[]for _ in range(n_cluster)]
#对样本遍历进行分类
for sample_point in samples:
dis=[cal_distance(sample_point,cluster_point) for cluster_point in cluster]
min_dis=min(dis)
index_cluster=dis.index(min_dis)
lists[index_cluster].append(sample_point)
#print(dis)

print(lists)
#设置不同聚类点的颜色
color_names=list(m_colors.cnames)

for i in lists:
color_=[color_names[lists.index(i)]]*len(i)
x=[]
y=[]
for classify_point in i:
x.append(classify_point[0])
y.append(classify_point[1])
plt.scatter(x,y,c=color_)
#重新计算聚集点中心
new_cluster.append([sum(x)/len(x),sum(y)/len(y)])
#对新产生的聚类中心作图
for cluster_point in new_cluster:
plt.scatter(cluster_point[0], cluster_point[1], c='red')
plt.title('output of k_means(iteration:{})'.format(n_loop))
plt.show()
print(len(cluster),len(new_cluster))
#计算新旧聚类中心的偏移距离
for i in range(len(cluster)):
print(i)
cutoff_dis[i]=cal_distance(cluster[i],new_cluster[i])

print('cluster',cluster)
print('new_cluster:',new_cluster)
print('cutoff_dis:',cutoff_dis)
#对聚类中心重新赋值
for i in range(len(cluster)):
cluster[i]=new_cluster[i]

new_cluster.clear()
n_loop+=1