机器学习实战-KNN算法

import numpy
import os
from numpy import array
from numpy import tile
import operator
import matplotlib.pyplot as plt

#数据例子
def createDataSet():
    group=array([[1.0,1.1],[1.0,1.0],[0,0],[0,0.1]])
    labels=['A','A','B','B'] #标签与点一一对应
    return group,labels
'''******************************主要分类函数************************************************************'''
#取距离最近的k各点，  返回 k个点中频率最多的类别作为分类
def classify0(point,dataArray,labels,k):#（测试[,,...]  比较集array  标签集  
    OneDimension=dataArray.shape[0]
    tmpArray=(tile(point,(OneDimension,1))-dataArray)**2 #point平铺成二维矩与其计算各点距离
    sqrtArray=tmpArray.sum(1)
    sortedArrayIndex=sqrtArray.argsort()#按索引点排序 -列表
    #print(sortedArrayIndex)
    classCount={} #空字典
    for i in range(k):
        lab=labels[sortedArrayIndex[i]]  #取相应索引点的标签
        classCount[lab]=classCount.get(lab,0)+1 #字典中有该key则取其映射值（这里为int），否则返回0
    sortedClassCount=sorted(classCount.items(),key=operator.itemgetter(1),reverse=True)
                                           #指明关键字
    return sortedClassCount[0][0]
#测试
'''
group0,labels0=createDataSet()
print(group0,labels0)
print(classify0([0,0],group0,labels0,3))
'''

'''*********************************************************************'''
'''***************************约会配对分类*******************************'''
'''*********************************************************************'''
#获取文件数据 返回数据Array 标签list
def file2matrix(filename):
    file=open(filename)
    fileList=file.readlines()#返回全部行 ，行后有\n---列表
    returnMat=numpy.zeros((len(fileList),3))
    index=0
    labels=[]
    for st in fileList:
        st=st.strip()#移除字符串头尾指定的字符（默认为空字符）  这里移除\n
        strList=st.split('\t')#str.split(sep=None, maxsplit=-1 无限制)
        returnMat[index,:]=strList[0:3]
        labels.append(int(strList[-1]))
        index+=1
    return returnMat,labels

#数据归一化
def Normalize(dataMat): #Array
    min_value=dataMat.min(0)
    max_value=dataMat.max(0)
    range_value=max_value-min_value
    normMat=dataMat-tile(min_value,(dataMat.shape[0],1))
    normMat=normMat/tile(range_value,(dataMat.shape[0],1))
    return normMat,range_value,min_value

#测试KNN错误率
def datingClassTest():
    datingData,datingLabels=file2matrix('datingTestSet2.txt')
    datingData,datingRange,datingMinValue=Normalize(datingData)
    testnum=int(datingData.shape[0]/10) #100
    error_count=0;
    for i in range(testnum):
        label=classify0(datingData[i],datingData[testnum:datingData.shape[0]],datingLabels[testnum:datingData.shape[0]],3)
        if label!=datingLabels[i]:
            error_count+=1;
    print('错误率:%f'%(error_count/float(testnum)))
        
    
#测试
'''
datingDataMat,datingLabels=file2matrix('datingTestSet2.txt')
datingDataMat,datingDataRange,datingDataMin=Normalize(datingDataMat)
fg=plt.figure()
subfg1=fg.add_subplot(111)
subfg1.scatter(datingDataMat[:,0],datingDataMat[:,1],15.0*array(datingLabels),15.0*array(datingLabels))
#subfg1.scatter(datingDataMat[:,0],datingDataMat[:,1],15.0*array(datingLabels),1*array(tile([1],(array(datingLabels).shape[0],1))))
plt.xlabel('玩视频耗时百分比')
plt.ylabel('周消耗冰激凌公升数')
plt.show()
datingClassTest()
'''
'''*********************************************************************'''
def classifyPerson():
    datingData,datingLabels=file2matrix('datingTestSet2.txt')
    datingData,datingRange,datingMinValue=Normalize(datingData)
    resultClass=['不喜欢','一般','有魅力']
    miles=float(input('每年飞行里程数：'))
    game=float(input('玩游戏小号百分比:'))
    ice=float(input('每周冰淇淋公升：'))
    data=array(([miles,game,ice]-datingMinValue)/datingRange)
    label=classify0(data,datingData,datingLabels,3)
    print('类型是:',resultClass[label-1]) #数据的分类标签1,2,3
    
#测试  
'''
classifyPerson()
'''

'''*********************************************************************'''
'''*****************************手写识别*********************************'''
'''*********************************************************************'''
def img2vector(filename):
    file=open(filename)
    returnVec=numpy.zeros((1,1024))
    for i in range(32):
        fileString=file.readline()
        for j in range(32):
            returnVec[0,i*32+j]=fileString[j]
    file.close()
    return returnVec

def handWriteClassTest():
    trainList=os.listdir('trainingDigits')
    DT=len(trainList)
    trainArray=numpy.zeros((DT,1024))
    labels=[]
    for i in range(DT):                                                    
        filename=trainList[i]
        labels.append(int(filename[0]))
        trainArray[i,:]=img2vector('trainingDigits/%s'%filename)
    testList=os.listdir('trainingDigits')
    DS=len(testList)
    error_count=0;
    for j in range(DS):
        filename=testList[j]
        label=int(filename[0])
        testArray=img2vector('trainingDigits/%s'%filename)
        testLabel=classify0(testArray,trainArray,labels,3)
        if label!=testLabel:
            error_count+=1
    error_rate=error_count/DS
    print('错误率：%f'%error_rate)
        
#测试       
'''
handWriteClassTest()
'''