python实现k_means聚类
K-Means算法是将一组N个样本的特征矩阵X划分为K个无交集的簇，直观上来看是簇是一组一组聚集在一起的数据，在一个簇中的数据就认为是同一类。簇就是聚类的结果表现。
簇中所有数据的均值通常被称为这个簇的“质心”(Centroids)。在一个二维平面中，一簇数据点的质心的横坐标就是这一簇数据点的横坐标的均值，质心的纵坐标就是这一簇数据点的纵坐标的均值。同理可推广至高维空间。
算法实现：
一、随机取n个点，即想要分类数目
二、将当前数据每个点都与取到的n个点进行作差，将差值最小的分为一类
三、将分好类的结果取出计算均值，取出每个类中和均值距离最小的值，将其做为新的中心点
四、重复上述步骤，直到最终分成的簇内结果不再发生变化（在当前的算法里，我们连续验证三次，即在三次内，结果不再发生变化，迭代就会停止）
#k均值计算
import numpy as np
import pandas as pd
import random
import math
def data():
rand_value = [random.randint(1,10000) for i in range(10000)]
# print(rand_value)
# rand_value = [63, 23, 93, 52, 33, 84, 34, 91, 52, 68, 64, 4, 3, 58, 37, 76, 75, 52, 49, 34, 91, 29, 67, 42, 97, 69, 99, 9, 15,
# 44, 12, 46, 86, 92, 10, 67, 46, 93, 49, 29, 93, 95, 60, 33, 30, 63, 24, 45, 26, 47, 12, 62, 50, 54, 13, 62, 40, 39,
# 32, 2, 92, 34, 20, 72, 68, 37, 88, 32, 84, 28, 88, 85, 5, 88, 56, 30, 5, 3, 43, 9, 13, 62, 44, 11, 90, 97, 80, 34,
# 12, 90, 56, 54, 87, 59, 20, 51, 58, 54, 29, 32]
return rand_value
#进行计算，并返回质心及详细信息
def k_mean(data_list,k,rand_choice):
# rand_choice = [random.choice(data_list) for i in range(k)]
temp_dict = {}
for i in range(len(data_list)):
temp_list = []
for j in range(len(rand_choice)):
#解决符号差异带来的影响
temp_list.append(abs(abs(data_list[i]) - abs(rand_choice[j])))
insert_index = temp_list.index(min(temp_list))
if temp_dict.get(rand_choice[insert_index]):
temp_dict[rand_choice[insert_index]].append(data_list[i])
else:
temp_dict[rand_choice[insert_index]] = [data_list[i]]
return temp_dict
#输出距离均值最近的点，返回新的质心及中心点
def min_distince(data_dict):
k_list = []
k_keys = list(data_dict.keys())
for j in range(len(k_keys)):
data_list = data_dict[k_keys[j]]
mean_value = sum(data_list)/len(data_list)
for i in range(len(data_list)):
if i == 0:
distance_value = abs(data_list[i] - mean_value)
s = 0
# distance_list.append(distance_value)
else:
if distance_value > abs(data_list[i] - mean_value):
s = i
distance_value = min(distance_value,abs(data_list[i] - mean_value))
k_list.append(data_list[s])
# distance_list.append(distance_value)
return k_list
def iteration(source_data,k):
iter_num = 0
end_list = []
if len(source_data) < k:
print('需要计算的数据量少于要分类的数据量，请检查数据源\n')
else:
while True:
iter_num += 1
print('迭代次数%d'%iter_num)
if iter_num == 1:
data_dict = k_mean(source_data,k,[random.choice(source_data) for i in range(k)])
end_list.append(data_dict)
# print(data_dict)
result = min_distince(data_dict)
# result = min_distince(k_mean(source_data,k,[random.choice(source_data) for i in range(k)]))
else:
data_dict = k_mean(source_data,k,result)
end_list.append(data_dict)
# print(data_dict)
result = min_distince(data_dict)
if len(end_list) == 4:
del end_list[0]
if end_list[0] == end_list[1] and end_list[1] == end_list[2]:
print(data_dict)
break
if __name__ == '__main__':
source_data = data()
iteration(source_data,5)