python实现k-means算法

python实现k-means算法

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import random
import math

data = pd.read_csv("D:/download/machine-learning/iris_data.csv")
# print(data)
datax = data[['Sepal.Length','Sepal.Width','Petal.Length','Petal.Width']]
data_X = np.array(datax.values)
lable = np.zeros(len(data_X))

#初始化3个聚类点
k1 = random.randint(0,50)
k2 = random.randint(50,100)
k3 = random.randint(100,150)

kmeans = []
kmeans.append(data_X[k1])
kmeans.append(data_X[k2])
kmeans.append(data_X[k3])
k1 = 0
k2 = 1
k3 = 2
#迭代次数
inter = 20

#计算平方和
def counter(i,k):
    sum = 0
    for j in range(4):
        sum += math.pow(data_X[i][j]- kmeans[k][j],2)
    return sum

for k in range(inter):
    print(kmeans)
    #计算各点到聚类中心点的距离
    for i in range(len(data_X)):
        i1 = math.sqrt(counter(i,k1))
        i2 = math.sqrt(counter(i,k2))
        i3 = math.sqrt(counter(i,k3))
        if i1<i2:
            if i1<i3: lable[i] = 1
            else: lable[i] = 3
        else:
            if i2<i3: lable[i] = 2
            else:lable[i] = 3
    #更新聚类中心点
    temp_k1 = np.zeros(4)
    temp_k2 = np.zeros(4)
    temp_k3 = np.zeros(4)
    c1 = 0
    c2 = 0
    c3 = 0
    for i in range(len(data_X)):
        if lable[i] == 1:
            temp_k1 += data_X[i]
            c1 += 1
        elif lable[i] == 2:
            temp_k2 += data_X[i]
            c2 += 1
        else: 
            temp_k3 += data_X[i]
            c3 += 1
    temp_k1 = temp_k1/c1
    temp_k2 = temp_k2/c2
    temp_k3 = temp_k3/c3
    kmeans[0] = temp_k1
    kmeans[1] = temp_k2
    kmeans[2] = temp_k3

#数据转换，为画图方便
print(type(str(lable[4])))
lable_str = []
for i in range(len(data_X)):
    lable_str.append(str(lable[i]))

datax['lable'] = lable_str

pd.plotting.andrews_curves(datax, 'lable')
plt.show()

实验课写的，可能有错误，欢迎指正。