k近邻算法探索总结

import requests
import pandas as pd
from numpy import *
import operator

# indices n. 指数；目录（index的复数）['ɪndɪsiːz]
def createDataSet():
    group=array([[1.0,1.1],[1.0,1.0],[0,0],[0,0.1]])
    label = ['A','A','B','B']
    return group,label

def classify0(inX,dataSet,labels,k):
    dataSetSize = dataSet.shape[0]
    diffMat = tile(inX,(dataSetSize,1))-dataSet
    sqDiffMat = diffMat**2
    # print(sqDiffMat)
    # axis =0是列相加，axis =1是行相加
    sqDistances = sqDiffMat.sum(axis=1)
    distances = sqDistances**0.5
    # print(distances)
    # argsort()先显示下标再从小到大排序，然后根据排序结果排序下标（注释1）
    sorteDistIndicies = distances.argsort()
    # print(sorteDistIndicies)
    classCount = {}
    for i in range(k):
        print(11111111111,sorteDistIndicies[i])
        voteIlabel = labels[sorteDistIndicies[i]]
        classCount[voteIlabel] = classCount.get(voteIlabel,0)+1
    # print(classCount)
    # 下面的sorted函数报错
    # 在python3报错，将iteritems改为items即可
    # iteritems是为python2环境中dict的函数
    # operator.itemgetter(1)报错,
    # 这个东西需要引用operator这个模块才可以使用这里面的1代表获取对象第一个域的值（注释2）
    # 这下面的key不能省略，reverse要注意没有d不是reversed
    sortedClassCount = sorted(classCount.items(),key=operator.itemgetter(1),reverse=True)
    # print(sortedClassCount)
    return sortedClassCount[0][0]

group,label = createDataSet()
last = classify0([0,0],group,label,3)
print(last)
# （注释1）argsort()
# [ 1.48660687  1.41421356  0.          0.1       ]
#     0              1        2           3
# [0. 0.1 1.41 1.48]
# 2 3 1 0

# （注释2）
# a = [1,2,3] 
# >>> b=operator.itemgetter(1)      //定义函数b，获取对象的第1个域的值
# >>> b(a) 
# 2 
# >>> b=operator.itemgetter(1,0)   //定义函数b，获取对象的第1个域和第0个的值
# >>> b(a) 
# (2, 1) 
# 来源https://blog.csdn.net/dongtingzhizi/article/details/12068205
# 第一小节，代码照着一遍，尝试默写一遍，默写一遍，共三遍

上面代码主要为创建测试数据以及分类器

# matrix    英[ˈmeɪtrɪks]
import pandas as pd
from numpy import *

def file2matrix(filename):
    fr = open(filename)
    arrayOLines = fr.readlines()
    numOfLines = len(arrayOLines)
    returnMat = zeros((numOfLines,3))
    # print(returnMat)
    classLabelVector = []
    index=0
    for line in arrayOLines:
        line = line.strip()
        listFromLine = line.split('\t')
        # print(listFromLine)
        returnMat[index,:] = listFromLine[:3]
        # print(returnMat)
        classLabelVector.append(int(listFromLine[-1]))
        # print(classLabelVector)
        index+=1
    return returnMat,classLabelVector


group,labels = file2matrix('datingTestSet2.txt')
# print(group,labels)
import matplotlib
import matplotlib.pyplot as plt
# n.    (代表数量，尤指官方资料中的) 数字; 数字符号; 字码; 位数; 算术;
fig = plt.figure()
# 这些是作为单个整数编码的子绘图网格参数。例如，“111”表示“1×1网格，第一子图”，“234”表示“2×3网格，第四子图”。(注释1)
ax = fig.add_subplot(111)
# print(group[:10])
# print(group[:10,0])
# print(group[:10,2])
# print(group[:10,1])
# scatter    英[ˈskætə(r)]    撒; 撒播; 散开; 四散; 使分散; 驱散;
# 前2个参数分别是他的x,y两个轴的数据，第三个是管点的大小的，第四个是管颜色的（注释2）
#scatter（x,y,s=1,c="g",marker="s",linewidths=0）
#s:散列点的大小,c:散列点的颜色，marker：形状，linewidths：边框宽度
# 目前没有看到加图例的方法，网上有2种，参考网址先种草（注释3）
ax.scatter(group[:,0],group[:,1],15.0*array(labels),15.0*array(labels),label='不喜欢')
plt.show()

# （注释1）https://www.jianshu.com/p/7b68e01952b4
# （注释2）https://www.cnblogs.com/shanlizi/p/6850318.html
# （注释3）https://www.jianshu.com/p/71cc65f9c30a
# （注释3）https://blog.csdn.net/xiaobaicai4552/article/details/79069207

上面代码为P21-P24内容探索

def autoNorm(group):
    k = group.shape[0]
    minVals = tile(group.min(0),(k,1))
    maxVals = tile(group.max(0),(k,1))
    normVals = group-minVals
    ranges = maxVals-minVals
    lastVals = normVals/ranges
    return lastVals,ranges,minVals

上面为P25归一化代码比较简单主要是套用公式

newValue = (oldValue-min)/(max-min)

def datingClassTest():
    testRate = 0.1
    errorCount = 0
    group, labels = createDataSet('datingTestSet2.txt')
    lastVals,ranges,minVals = autoNorm(group)
    testNum = int(testRate*lastVals.shape[0])
    print(testNum)
    for i in range(testNum):
        classfyVals = classify0(lastVals[i,:],lastVals[testNum:,:],labels[testNum:],3)
        print(classfyVals)
        if (classfyVals!=labels[i]):errorCount+=1
    print('the errorrate is %f'%(errorCount/testNum))

上面部分为约会网站最终测试分类成功率的代码在书里P27

下面代码为P28页完整使用各种方法分辨手写数字的函数，由于是自己写的可能和书里的有点不一样，但是大体原理是一致的

import pandas as pd
from numpy import *
import operator
import os
def createDataSet(path):
    op = open(path,'r')
    baseSet = op.readlines()
    nullSet = zeros((len(baseSet), 3))
    index=0
    labelsList = []
    for i in baseSet:
        baseList = i.strip().split('\t')
        # print(baseList)
        nullSet[index,:] = baseList[:3]
        labelsList.append(baseList[-1])
        index+=1
    return nullSet,labelsList

def autoNum(dataSet):
    minVals = tile(dataSet.min(0),(len(dataSet),1))
    maxVals = tile(dataSet.max(0),(len(dataSet),1))
    lastValue = (dataSet-minVals)/(maxVals-minVals)
    # print(lastValue)
    return lastValue

def classify0(inX,dataSet,labels,k):
    inXSet = tile(inX,(len(dataSet),1))
    distance = ((dataSet-inXSet)**2).sum(axis=1)**0.5
    listIndex = distance.argsort()
    # print(listIndex)
    classDict = {}
    for i in range(int(k)):
        classDict[labels[listIndex[i]]] = classDict.get(labels[listIndex[i]],0)+1

    lastSet = sorted(classDict.items(),key=operator.itemgetter(1),reverse=True)
    return lastSet[0][0]

# 这个是约会网站的那个案例的分类测试
def classifyPerson():
    group,labels = createDataSet('datingTestSet2.txt')
    lastValue = autoNum(group)
    destRate = 0.1
    errorCount = 0
    destNum = int(destRate*len(group))
    # print(destNum)
    for i in range(destNum):
        choiceLabel = classify0(group[i,:],group[destNum:],labels[destNum:],3)
        print(choiceLabel)
        if choiceLabel!=labels[i]:
            errorCount+=1
    print('the error rate is %f'%(errorCount/destNum))
# 从这个函数开始，是P28页开始的手写识别系统的代码
def img2vector(path):
    op = open(path,'r')
    dataSet = op.readlines()
    # print(dataSet)
    nullSet = zeros((1,1024))
    for i in range(32):
        for j in range(32):
            nullSet[0,i*32+j] = dataSet[i][j]
    return nullSet

def testWrite(path,testpath):
    fileNames = os.listdir(path)
    nullSet = zeros((len(fileNames),1024))
    labels = []
    for i in range(len(fileNames)):
        fileLabels = fileNames[i].split('_')[0]
        nullSet[i,:] = img2vector(path+'\\%s'%fileNames[i])
        labels.append(fileLabels)

    testFileName = os.listdir(testpath)
    errorCount = 0
    for ii in testFileName:
        testData = img2vector(testpath + '\\%s' % ii)
        RightLabel = ii.split('_')[0]
        classifyVals = classify0(testData,nullSet,labels,3)
        print(RightLabel,classifyVals)
        if RightLabel!=classifyVals:
            errorCount+=1
    print('the test error rate is %f'%(errorCount/len(testFileName)))

testWrite('E:\\10机器学习\\Digits\\trainingDigits','E:\\10机器学习\\Digits\\testDigits')