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)