机器学习_K近邻Python代码详解

k近邻优点：精度高、对异常值不敏感、无数据输入假定；
k近邻缺点：计算复杂度高、空间复杂度高

import numpy as np
import operator
from os import listdir

# k近邻分类器
def classify0(inx, dataSet, labels, k):
    dataSetSize = dataSet.shape[0]   # 返回dataset第一维的长度，也就是行数
    diffMat = np.tile(inx, (dataSetSize, 1))-dataSet   # tile表示把inx行向量按列方向重复datasetsize次
    sqDiffMat = diffMat**2
    sqDistances = sqDiffMat.sum(axis=1)  # 按列求和
    distances = sqDistances**0.5
    sortedDistIndicies = distances.argsort()  # 返回的是数组从小到大的索引值
    classCount = {}    # 定义一个空字典
    for i in range(k):
        voteLabel = labels[sortedDistIndicies[i]]  # 返回前k个距离最小的样本的标签值
        classCount[voteLabel] = classCount.get(voteLabel, 0)+1   # get 表示返回指定键的值
        # lambda表示输入classCount返回冒号右边的值，reverse=True表示按照降序排列
        sortedClassCount=sorted(classCount.items(), key=lambda classCount: classCount[1], reverse=True)
    return sortedClassCount[0][0]

# 把.txt文件转换成矩阵形式
def file2matrix(file):
    file = open(file) # 返回文件对象
    arr = file.readlines()  # 返回全部行，是list形式，每一行为list的一个元素
    number = len(arr)  # 返回对象长度
    returnMat = np.zeros((number,3))
    index = 0
    labelMat = []
    for line in arr:
        #line = line.strip('\n')
        #newline = line.split(' ')
        newline = line.strip('\n').split(' ') # 处理逐行数据，strip表示把头尾的'\n'去掉，split表示以空格来分割行数据
                                              #  然后把处理后的行数据返回到newline列表中
        returnMat[index,:] = newline[0:3] #表示列表的0,1,2列数据放到index行中
        labelMat.append(int(newline[-1]))
        index+=1
    return returnMat,labelMat

# 归一化
def autoNorm(dataSet):
    minVals = dataSet.min(0)

    maxVals = dataSet.max(0)
    ranges = maxVals-minVals
    normDataSet = np.zeros(np.shape(dataSet))
    m = normDataSet.shape[0]
    A = normDataSet
    A = np.tile(minVals, (m,1))
    normDataSet = dataSet-A
    normDataSet = normDataSet/np.tile(ranges,(m,1))
    return normDataSet

# 把图像转化成向量的形式
def img2vector(filename):
    returnVect = np.zeros((1,1024))
    fr = open(filename)
    for i in range(32):
        lineStr = fr.readline() # readline()表示从首行开始，每次读取一行
        for j in range(32):
            returnVect[0,32*i+j] = int(lineStr[j]) #int()函数用于将一个字符串或数字转换成整型
    return returnVect   # 一张图片转化成一行后的数组

# 手写数字识别系统的测试代码
def handwritingClassTest():
    hwLabels = []
    trainingFileList = listdir('E:/workspace/digits/trainingDigits')
    m=len(trainingFileList)
    trainingMat = np.zeros((m,1024))
    for i in range(m):
        fileNameStr = trainingFileList[i]  # 例如9_45.txt
        fileStr = fileNameStr.split('.')[0]  # split('.')通过.分隔符对字符串进行切片
        classNumStr = int(fileStr.split('_')[0]) # split('_')通过_分隔符对字符串进行切片
        hwLabels.append(classNumStr)
        trainingMat[i,:] =img2vector('E:/workspace/digits/trainingDigits/%s' % fileNameStr)
    testFileList = listdir('E:/workspace/digits/testDigits')
    mTest = len(testFileList)
    errorCount = 0
    for i in range(mTest):
        fileNameStr = testFileList[i]
        fileStr = fileNameStr.split('.')[0]
        classNumStr = int(fileStr.split('_')[0])
        vectorUnderTest = img2vector('E:/workspace/digits/testDigits/%s' % fileNameStr)
        classResult = classify0(vectorUnderTest,trainingMat,hwLabels,3)
        print('the classifier came back with: %d, the real answer is: %d' % (classResult,classNumStr))
        if (classResult != classNumStr):
            errorCount += 1.0
    print('\n the total number of errors is: %d' % (errorCount))
    print('\n the total error rate is: %f' % (errorCount/float(mTest)))

handwritingClassTest()