训练样本是多维特征空间向量,当中每一个训练样本带有一个类别标签(喜欢或者不喜欢、保留或者删除)。分类算法常採用“多数表决”决定,即k个邻居中出现次数最多的那个类作为预測类。“多数表决”分类的一个缺点是出现频率较多的样本将会主导測试点的预測结果。那是由于他们比較大可能出如今測试点的K邻域而測试点的属性又是通过K领域内的样本计算出来的。解决这个缺点的方法之中的一个是在进行分类时将K个邻居到測试点的距离考虑进去。比如,若样本到測试点距离为d,则选1/d为该邻居的权重(也就是得到了该邻居所属类的权重),接下来统计统计k个邻居全部类标签的权重和,值最大的那个就是新数据点的预測类标签。
举例,K=5,计算出新数据点到近期的五个邻居的举例是(1,3,3,4,5),五个邻居的类标签是(yes,no,no。yes,no)
若是依照多数表决法,则新数据点类别为no(3个no,2个yes);若考虑距离权重类别则为yes(no:2/3+1/5,yes:1+1/4)。
以下的Python程序是採用KNN算法的实例(计算欧氏距离。多数表决法决断):一个是採用KNN算法改进约会站点配对效果。还有一个是採用KNN算法进行手写识别。
约会站点配对效果改进的样例是依据男子的每年的飞行里程、视频游戏时间比和每周冰激凌耗量三个特征来推断其是否是海伦姑娘喜欢的类型(类别为非常喜欢、一般和讨厌),决策採用多数表决法。
因为三个特征的取值范围不同,这里採用的scale策略为归一化。
使用KNN分类器的手写识别系统 仅仅能识别数字0到9。
须要识别的数字使用图形处理软件,处理成具有同样的色 彩和大小 :宽髙是32像素X32像素的黑白图像。虽然採用文本格式存储图像不能有效地利用内存空间,为了方便理解,这里已经将将图像转换为文本格式。训练数据中每一个数字大概有200个样本。程序中将图像样本格式化处理为向量,即一个把一个32x32的二进制图像矩阵转换为一个1x1024的向量。from numpy import *
import operator
from os import listdir
import matplotlib
import matplotlib.pyplot as plt
import pdb
def classify0(inX, dataSet, labels, k=3):
#pdb.set_trace()
dataSetSize = dataSet.shape[0]
diffMat = tile(inX, (dataSetSize,1)) - dataSet
sqDiffMat = diffMat**2
sqDistances = sqDiffMat.sum(axis=1)
distances = sqDistances**0.5
sortedDistIndicies = distances.argsort() #ascend sorted,
#return the index of unsorted, that is to choose the least 3 item
classCount={}
for i in range(k):
voteIlabel = labels[sortedDistIndicies[i]]
classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1# a dict with label as key and occurrence number as value
sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True)
'''descend sorted according to value, '''
return sortedClassCount[0][0]
def file2matrix(filename):
fr = open(filename)
#pdb.set_trace()
L = fr.readlines()
numberOfLines = len(L) #get the number of lines in the file
returnMat = zeros((numberOfLines,3)) #prepare matrix to return
classLabelVector = [] #prepare labels return
index = 0
for line in L:
line = line.strip()
listFromLine = line.split('\t')
returnMat[index,:] = listFromLine[0:3]
classLabelVector.append(int(listFromLine[-1]))
#classLabelVector.append((listFromLine[-1]))
index += 1
fr.close()
return returnMat,classLabelVector
def plotscattter():
datingDataMat,datingLabels = file2matrix('datingTestSet2.txt') #load data setfrom file
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax2 = fig.add_subplot(111)
ax3 = fig.add_subplot(111)
ax1.scatter(datingDataMat[:,0],datingDataMat[:,1],15.0*array(datingLabels),15.0*array(datingLabels))
#ax2.scatter(datingDataMat[:,0],datingDataMat[:,2],15.0*array(datingLabels),15.0*array(datingLabels))
#ax2.scatter(datingDataMat[:,1],datingDataMat[:,2],15.0*array(datingLabels),15.0*array(datingLabels))
plt.show()
def autoNorm(dataSet):
minVals = dataSet.min(0)
maxVals = dataSet.max(0)
ranges = maxVals - minVals
normDataSet = zeros(shape(dataSet))
m = dataSet.shape[0]
normDataSet = dataSet - tile(minVals, (m,1))
normDataSet = normDataSet/tile(ranges, (m,1)) #element wise divide
return normDataSet, ranges, minVals
def datingClassTest(hoRatio = 0.20):
#hold out 10%
datingDataMat,datingLabels = file2matrix('datingTestSet2.txt') #load data setfrom file
normMat, ranges, minVals = autoNorm(datingDataMat)
m = normMat.shape[0]
numTestVecs = int(m*hoRatio)
errorCount = 0.0
for i in range(numTestVecs):
classifierResult = classify0(normMat[i,:],normMat[numTestVecs:m,:],datingLabels[numTestVecs:m],3)
print "the classifier came back with: %d, the real answer is: %d" % (classifierResult, datingLabels[i])
if (classifierResult != datingLabels[i]): errorCount += 1.0
print "the total error rate is: %.2f%%" % (100*errorCount/float(numTestVecs))
print 'testcount is %s, errorCount is %s' %(numTestVecs,errorCount)
def classifyPerson():
'''
input a person , decide like or not, then update the DB
'''
resultlist = ['not at all','little doses','large doses']
percentTats = float(raw_input('input the person\' percentage of time playing video games:'))
ffMiles = float(raw_input('flier miles in a year:'))
iceCream = float(raw_input('amount of iceCream consumed per year:'))
datingDataMat,datingLabels = file2matrix('datingTestSet2.txt')
normMat, ranges, minVals = autoNorm(datingDataMat)
normPerson = (array([ffMiles,percentTats,iceCream])-minVals)/ranges
result = classify0(normPerson, normMat, datingLabels, 3)
print 'you will probably like this guy in:', resultlist[result -1]
#update the datingTestSet
print 'update dating DB'
tmp = '\t'.join([repr(ffMiles),repr(percentTats),repr(iceCream),repr(result)])+'\n'
with open('datingTestSet2.txt','a') as fr:
fr.write(tmp)
def img2file(filename):
#vector = zeros(1,1024)
with open(filename) as fr:
L=fr.readlines()
vector =[int(L[i][j]) for i in range(32) for j in range(32)]
return array(vector,dtype = float)
def handwritingClassTest():
hwLabels = []
trainingFileList = listdir('trainingDigits') #load the training set
m = len(trainingFileList)
trainingMat = zeros((m,1024))
for i in range(m):
fileNameStr = trainingFileList[i]
fileStr = fileNameStr.split('.')[0] #take off .txt
classNumStr = int(fileStr.split('_')[0])
hwLabels.append(classNumStr)
trainingMat[i,:] = img2vector('trainingDigits/%s' % fileNameStr)
testFileList = listdir('testDigits') #iterate through the test set
errorCount = 0.0
mTest = len(testFileList)
for i in range(mTest):
fileNameStr = testFileList[i]
fileStr = fileNameStr.split('.')[0] #take off .txt
classNumStr = int(fileStr.split('_')[0])
vectorUnderTest = img2vector('testDigits/%s' % fileNameStr)
classifierResult = classify0(vectorUnderTest, trainingMat, hwLabels, 3)
print "the classifier came back with: %d, the real answer is: %d" % (classifierResult, classNumStr)
if (classifierResult != classNumStr): errorCount += 1.0
print "\nthe total number of errors is: %d" % errorCount
print "\nthe total error rate is: %f" % (errorCount/float(mTest))
if __name__ == '__main__':
datingClassTest()
#handwritingClassTest()
KNN算法学习包下载地址为:
(四)KNN回归
数据点的类别标签是连续值时应用KNN算法就是回归。与KNN分类算法过程同样。差别在于对K个邻居的处理上。KNN回归是取K个邻居类标签值得加权作为新数据点的预測值。加权方法有:K个近邻的属性值的平均值(最差)、1/d为权重(有效的衡量邻居的权重。使较近邻居的权重比較远邻居的权重大)、高斯函数(或者其它适当的减函数)计算权重= gaussian(distance) (距离越远得到的值就越小,加权得到更为准确的预计。
(五)总结
K-近邻算法是分类数据最简单最有效的算法,其学习基于实例,使用算法时我们必须有接近实际数据的训练样本数据。K-近邻算法必须保存所有数据集,假设训练数据集的非常大,必须使用大量的存储空间。此外,因为必须对数据集中的每一个数据计算距离值,实际使用时可能非常耗时。
k-近邻算法的还有一个缺陷是它无法给出不论什么数据的基础结构信息,因此我们也无法知晓平均实例样本和典型实例样本具有什么特征。