numpy实现k近邻法

import numpy as np 

import matplotlib.pyplot as plt  

w=250

train=np.random.randint(-300,300,(w,4))

train=train.astype(float)

train_lable=np.zeros(w)

def kzhe():

    w1=np.zeros((50,4))

    w1_lable=np.zeros(50)

    for i in range(50):

        w1[i]=train[i+200]

        w1_lable[i]=train_lable[i+200]

    i=3

    for j in range(4):

        train[50*(i+1):50*(i+2),:]=train[50*i:50*(i+1),:]

        train_lable[50*(i+1):50*(i+2)] =train_lable[50*i:50*(i+1)]

        i=i-1

    train[0:50,:]=w1

    train_lable[0:50]=w1_lable  

for i in range(4):

    train[:,i]=(train[:,i]-train[:,i].mean())/train[:,i].std()

for i in range(w):

    if 1*train[i,0]+2*train[i,1]+3*train[i,2]+4*train[i,3]-1>0:

        train_lable[i]=1

    else:

        train_lable[i]=-1

def knn(index,k,sig):

    dis=np.zeros(200)

    disl=[]

    for i in range(200):

        dis[i]=(train[index,0]-train[i,0])*(train[index,0]-train[i,0])+(train[index,1]-train[i,1])*(train[index,1]-train[i,1])+(train[index,2]-train[i,2])*(train[index,2]-train[i,2])+(train[index,3]-train[i,3])*(train[index,3]-train[i,3])

        disl.append(dis[i])

    target=[]

    weight=[]

    for i in range(k):

        target.append(disl.index(min(disl)))

        weight.append(np.exp(-min(disl)**2/(2*sig**2)))

        disl[disl.index(min(disl))]=float('inf')

    l1=0

    l2=0

    for i in range(k):

        if train_lable[target[i]]==1:

            l1=l1+1*weight[i]

        else:

            l2=l2+1*weight[i]

    if l1>l2 :

        return 1

    else:

        return -1

best_score=-1

best_k=0

best_wid=0

for k in range(10):

    for wid in range(20):

        if k!=0 and wid !=0:

            sum2=0

            for j in range(5):

                kzhe()

                for i in range(50):

                    lable=knn(i+200,k,wid)

                    if lable==train_lable[i+200]:

                        sum2=sum2+1

                

            if best_score<sum2*1.0/250:

                best_score=sum2*1.0/250

                best_k=k

                best_wid=wid

            print("准确率",sum2*1.0/250,"k值：",k,"宽度:",wid)

print(best_score,best_k,best_wid)