numpy实现k均值聚类

import numpy as np 

import matplotlib.pyplot as plt

train=np.random.randint(-300,300,(300,5))

w1=np.random.randint(-100,100,(1,4))

def cir(data):

    max=-1

    for i in range(len(data)-1):

        for j in range(i+1,len(data)):

            if max<(data[i,0]-data[j,0])*(data[i,0]-data[j,0])+(data[i,1]-data[j,1])*(data[i,1]-data[j,1])+(data[i,2]-data[j,2])*(data[i,2]-data[j,2])+(data[i,3]-data[j,3])*(data[i,3]-data[j,3]):

                max=(data[i,0]-data[j,0])*(data[i,0]-data[j,0])+(data[i,1]-data[j,1])*(data[i,1]-data[j,1])+(data[i,2]-data[j,2])*(data[i,2]-data[j,2])+(data[i,3]-data[j,3])*(data[i,3]-data[j,3])

    return max

def k(i):

    w1=np.random.randint(-100,100,(i+1,4))

    w1=np.double(w1)

    w2=np.copy(w1)

    flag=1

    while 1:

        if flag==0 :

            break

        for i1 in range(len(train)):

            index=0

            min=float('inf')

            for j in range(i+1):

                if min>(train[i1,0]-w1[j,0])*(train[i1,0]-w1[j,0])+(train[i1,1]-w1[j,1])*(train[i1,1]-w1[j,1])+(train[i1,2]-w1[j,2])*(train[i1,2]-w1[j,2])+(train[i1,3]-w1[j,3])*(train[i1,3]-w1[j,3]):

                    min=(train[i1,0]-w1[j,0])*(train[i1,0]-w1[j,0])+(train[i1,1]-w1[j,1])*(train[i1,1]-w1[j,1])+(train[i1,2]-w1[j,2])*(train[i1,2]-w1[j,2])+(train[i1,3]-w1[j,3])*(train[i1,3]-w1[j,3])

                    index=j

            train[i1,4]=index           

        flag=0

        w2=np.copy(w1)        

        for j in range(i+1):

            w=[]

            for i1 in range(len(train)):

                if train[i1,4]==j:

                    w.append(train[i1])

            w=np.array(w)

            if len(w)!=0:

                w1[j,0]=w[0:,0].mean()

                w1[j,1]=w[0:,1].mean()

                w1[j,2]=w[0:,2].mean()

                w1[j,3]=w[0:,3].mean()              

        for j in range(i):

            for k3 in range(4):

                if w1[j,k3]!=w2[j,k3]:

                    flag=1

    loss=0

    if flag==0:

        for j in range(i+1):

            w4=[]

            for k1 in range(len(train)):

                if train[k1,4]==j:

                    w4.append(train[k1])

            w4=np.array(w4)

            loss=loss+cir(w4)

    return loss/(i+1)

zer=np.zeros(200)

x=[]

for i in range(len(zer)):

    zer[i]=k(i)

    x.append(i+1)

x=np.array(x)

plt.plot(x,zer)

print(zer)

plt.show()