KNN算法思想与实现

第二章 k近邻

2.1 算法描述

（1）采用测量不同特征值之间的距离进行分类

优点：对异常点不敏感，精度高，无数据输入设定

缺点：空间，计算复杂度高

适合数据：标称与数值

（2）算法的工作原理：

基于已有的带有标签的训练数据，计算出需要预测的数据与每个训练数据之间的距离，找到其中距离最近的k个数据，根据这k数据中数量最多的类来决定测试数据的类别

（3）算法的类别

该算法属于有监督学习，用于分类，因此它的目标变量是离散的

（4）算法的一般流程：

1.收集数据

2.准备数据

3.分析数据

4.测试算法

5.使用算法

2.2算法实现过程

（1）获取数据

　　（2）KNN算法

from numpy import *
import operator

# this KNN matrix col is 3
# in order to create data
def createDataSet():
    group = array([[1.0, 1.1], [1.0, 1.0], [0.0, 0.0], [0.0, 0.1]])
    lables = ['A', 'A', 'B', 'B']
    return group, lables

# main algorithm
def classify0(inx, dataSet, lables, k):
    datasetSize = dataSet.shape[0]
    diffmat = tile(inx, (datasetSize, 1)) - dataSet
    sqdiffmat = diffmat**2
    sqDistance = sqdiffmat.sum(axis=1)
    distance = sqDistance**0.5
    sortedDistance = distance.argsort()
    classcount = {}
    for i in range(k):
        votelabel = lables[sortedDistance[i]]
        classcount[votelabel] = classcount.get(votelabel, 0) + 1
    sortedclasscount = sorted(classcount.iteritems(), key=operator.itemgetter(1), reverse=True)
    return sortedclasscount[0][0]

# read the txt data file
def file2matrix(filename):
    fr = open(filename)
    arraylines = fr.readlines()
    numberoflines = len(arraylines)
    returnmatrix = zeros((numberoflines, 3))  # you can change the col
    clasavector = []
    index = 0
    for line in arraylines:
        line = line.strip()
        listformline = line.split('\t')
        returnmatrix[index, :] = listformline[0:3]  # you should change the col
        clasavector.append(int(listformline[-1]))
        index += 1
    return returnmatrix, clasavector

# normalize the data
def autonorm(dataset):
    minval = dataset.min(0)
    maxval = dataset.max(0)
    ranges = maxval - minval
    datasetsize = dataset.shape[0]
    normdataset = dataset - tile(minval, (datasetsize, 1))
    normdataset = normdataset/tile(ranges, (datasetsize, 1))
    return normdataset, ranges, minval

def datingclasstest(filename):
    horatio = 0.1
    dataset, lableset = file2matrix(filename)
    noramdataset, ranges, minval = autonorm(dataset)
    col = dataset.shape[0]
    test = int(col*horatio)
    errorcount = 0.0
    for i in range(col):
        classlable = classify0(noramdataset[i, :], noramdataset[test:col, :], lableset[test:col], 3)
        if classlable != lableset[i]:
            errorcount += 1
    error = errorcount / float(col)
    print error

　　（3）dating应用程序

import KNN
from numpy import *

def classifyperson():
    returnlist = ['not at all', 'in small doses', 'in large doses']
    game = float(raw_input("the percentage of playing video game"))
    fly = float(raw_input("the num of the flier mail"))
    icecream = float(raw_input("the num of icecream every weak"))
    person = array([game, fly, icecream])
    dataset,datalable =           KNN.file2matrix("F:data/machinelearninginaction/Ch02/datingTestSet2.txt")
    normdataset, ranges, minval=KNN.autonorm(dataset)
    classifierresult =KNN.classify0((person - minval)/ranges, normdataset, datalable, 3)
    print "you will like him %s" % returnlist[classifierresult-1]

　　（4）手写识别程序

import KNN
from os import listdir
from numpy import *

# change the 32*32 to vector
def image2vertor(filename):
    fr = open(filename)
    imagevertor = zeros((1, 1024))
    for i in range(32):
        line = fr.readline()
        for j in range(32):
            imagevertor[0, i*32+j] = int(line[j])
    return imagevertor
testvector = image2vertor("F:data/machinelearninginaction/Ch02/digits/testDigits/0_13.txt")

def handwritingtest():
    hwlables = []   # record the lable
    filename = listdir("F:data/machinelearninginaction/Ch02/digits/trainingDigits/")
    filenum = len(filename)
    dataset = zeros((filenum, 1024))
    for i in range(filenum):
        filenamestr = filename[i].split(".")[0]
        filelable = int(filenamestr.split('_')[0])
        hwlables.append(filelable)
        filepath = "F:data/machinelearninginaction/Ch02/digits/trainingDigits/" + filename[i]
        data = image2vertor(filepath)
        dataset[i, :] = data
    testfile = listdir("F:data/machinelearninginaction/Ch02/digits/testDigits/")
    testfilenum = len(testfile)
    for j in range(testfilenum):
        testfilestr = testfile[j].split('.')[0]
        testfilelable =int(testfilestr.split('_')[0])
        testdilepath = "F:data/machinelearninginaction/Ch02/digits/testDigits/" + testfile[j]
        testdata = image2vertor(testdilepath)
        classname = KNN.classify0(testdata, dataset, hwlables, 3)
        error = 0.0
        if classname == testfilelable:
            error += 1
        print "we think it is %d, the real is %d" % (classname, testfilelable)
    print "the num of error is %d " % error
    print "the error rate is %f" % (error/float(testfilenum))

handwritingtest()