一、准备工作
前面的博客已讲述了K近邻的概念,这里是一个K近邻算法的应用。
准备内容:
1.手写数字训练集trainingDigits
2.手写数字测试集testDigits
程序的基本内容:
1.原算法的第一步是导入数据,并进行矩阵化处理,这里的第一步是将图片转化成向量。图片在数据集中是以数字的形式保存,3232,将二维图片转化成11024一维的向量。一共有接近2000的训练集和900多的测试集
2.将每一个测试样本都调用KNN算法,KNN算法的参数为1.测试向量 2.训练样本 3.类别 4.k值
代码:
import numpy as np
import operator
import os
# 图像转换为向量
def img2vector(filename):
returnVect = np.zeros((1,1024)) # 生成一个1*1024的向量
fr = open(filename) # 打开文件
for i in range(32):
lineStr = fr.readline() # 一行一行的读
for j in range(32):
returnVect[0,32*i+j] = int(lineStr[j]) # 一个一个的加入1*1024的向量中
return returnVect # 返回这个1*1024向量
# KNN
def classify0(inX,dataset,labels,k):
dataSetSize = dataset.shape[0]
# 下面计算距离
diffMat = np.tile(inX,(dataSetSize,1))-dataset # tile()复制函数
sqDiffMat = diffMat**2
sqDistances = sqDiffMat.sum(axis=1) # 进行行相加
distances = sqDistances**0.5
sortedDistIndicies = distances.argsort() #argsort()按照从小到大排序返回原序列元素的序号。
# 选择距离最小的K个点
classCount={}
for i in range(k):
voteIlabel =labels[sortedDistIndicies[i]] # 找到与数据最近的k个标签
classCount[voteIlabel] = classCount.get(voteIlabel,0)+1 # 将各个标签的数量添加到字典中
# operator.itemgetter(1)选择字典的value进行排序,拍完序reserve反转
sortedClassCount = sorted(classCount.items(),key=operator.itemgetter(1),reverse=True)
# print(sortedClassCount)
return sortedClassCount[0][0]
# 测试手写数字识别
def handwritingClassTest():
hwlabels = []
trainingFileList = os.listdir('trainingDigits')
m = len(trainingFileList)
trainingMat = np.zeros((m,1024))
for i in range(m):
fileNameStr = trainingFileList[i] # 获取训练样本名字
fileStr = fileNameStr.split('.')[0] #获取训练样本的名字去掉后缀
classNumStr = int(fileStr.split('_')[0]) # 获取‘0_13’前面数字,代表数字
hwlabels.append(classNumStr) # 类别标签保存到列表中
trainingMat[i,:] = img2vector('trainingDigits/%s' %fileNameStr) # 将全部训练样本组成一个矩阵
testFileList = os.listdir('testDigits') # 获取测试样本的列表
errorCount = 0.0 # 错误分类
mTest = len(testFileList) # 测试样本的数量
for i in range(mTest):
fileNameStr = testFileList[i] # 获取测试样本的全部名称
fileStr = fileNameStr.split('.')[0] # 获取测试样本的名称
classNumStr = int(fileStr.split('_')[0]) # 得到类别名
vectorUnderTest = img2vector('testDigits/%s'%fileNameStr) # 转化成1*1024向量
classifierResult = classify0(vectorUnderTest,trainingMat,hwlabels,3) # 类别预测结果
print('the classifier came back with :%d,the real answer is :%d' %(classifierResult,classNumStr))
if classifierResult!=classNumStr:
errorCount += 1.0
print('the total number of error is : %d' %errorCount)
print('the total error rate is :%f' % (errorCount/float(mTest)))
handwritingClassTest()
结果:
the classifier came back with :0,the real answer is :0
the classifier came back with :0,the real answer is :0
the classifier came back with :0,the real answer is :0
the classifier came back with :0,the real answer is :0
the classifier came back with :0,the real answer is :0
the classifier came back with :0,the real answer is :0
.
.
the classifier came back with :2,the real answer is :2
the classifier came back with :2,the real answer is :2
the classifier came back with :2,the real answer is :2
the classifier came back with :2,the real answer is :2
the classifier came back with :2,the real answer is :2
the classifier came back with :3,the real answer is :3
the classifier came back with :3,the real answer is :3
the classifier came back with :3,the real answer is :3
.
.
the classifier came back with :9,the real answer is :9
the classifier came back with :9,the real answer is :9
the classifier came back with :9,the real answer is :9
the total number of error is : 11
the total error rate is :0.011628
Process finished with exit code 0
可以看到预测的错误率为1.1%。通过更改变量K值和修改函数的handwritingClassTest随机选取训练样本、改变训练样本的数目,都会对K近邻算法的错误率产生影响。
样本集下载:百度云链接
提取码:g64m