KNN

最新推荐文章于 2020-07-13 10:36:01 发布

原创最新推荐文章于 2020-07-13 10:36:01 发布 · 343 阅读

0 ·

CC 4.0 BY-SA版权

machine learning 专栏收录该内容

10 篇文章

订阅专栏

本文详细介绍了KNN算法的工作原理及其在手写数字识别中的应用。通过具体实例展示了如何使用KNN进行分类，并实现了从图像文件中加载手写数字数据集的过程。最后，通过实验验证了该算法的有效性。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

缺点：

耗内存，存储所有训练样本，对每个测试样本都要计算和所有训练数据的距离，时间成本高

knn 和 Locally weighted linear regression 思想上非常相似，对每个预测点都需要训练单独训练模型

代码如下：

from numpy import *
import os

def loaddata():
data=array([[1.0, 0.9], [1.0, 1.0], [0.1, 0.2], [0.0, 0.1]])
labels = ['A', 'A', 'B', 'B']
return data,labels

def img2vect(filename):
rows=32
cols=32
imgvect=zeros((1,rows*cols))
f=open(filename)
for i in range(rows):
linstr=f.readline()
for j in range(cols):
imgvect[0,i*rows+j]=int(linstr[j])
return imgvect

def loaddigit():
print('get training sets:')
filedir='C:/Users/yourname/Desktop/machin/digit/'
trainfilelist=os.listdir(filedir+'trainingDigits')
numsamples=len(trainfilelist)
train_x=zeros((numsamples,1024))
train_y=[]
for i in range(numsamples):
filename=trainfilelist[i]
train_x[i,:]=img2vect(filedir+'trainingDigits/%s' % filename)
label=int(filename.split('_')[0])
train_y.append(label)

print( "---Getting testing set..." )
testingFileList = os.listdir(filedir + 'testDigits')
numSamples = len(testingFileList)
test_x = zeros((numSamples, 1024))
test_y = []
for i in range(numSamples):
filename = testingFileList[i]
test_x[i, :] = img2vect(filedir + 'testDigits/%s' % filename)
label = int(filename.split('_')[0])
test_y.append(label)

return train_x, train_y, test_x, test_y

def knn(dataset,testdata,labels,k):
n=dataset.shape[0]
diff=tile(testdata,(n,1))-dataset
sdiff=diff**2
sumdist=sum(sdiff,axis=1)
dist=sumdist**0.5
dist_sort=argsort(dist)
classcount={}
for i in range(k):
label=labels[dist_sort[i]]
classcount[label]=classcount.get(label,0)+1
maxcount=0
for key,value in classcount.items():
if value>maxcount:
maxcount=value
maxindex=key
return maxindex

def testHandWritingClass():
train_x, train_y, test_x, test_y=loaddigit()
numtestsamples=test_x.shape[0]
matchcount=0
for i in range(numtestsamples):
predict=knn(train_x,test_x[i],train_y,3)
if predict==test_y[i]:
matchcount+=1
accuracy=float(matchcount)/numtestsamples
print('The classify accuracy is: %.2f%%' % (accuracy * 100))

data,labels=loaddata()
testdata=[0.8,0.7]
ke=knn(data,testdata,labels,2)
print(ke)
testHandWritingClass()