機(jī)器學(xué)習(xí)knn分類算法與應(yīng)用簡化版

KNNKNNKNNKNNKNNKNN分類算法PythonKNNKNNKNNKNNk1、理解KNN算法的思2KNNkNN本質(zhì)上，KNN3k找鄰居：圈定距離最近的k做分類：根據(jù)這k新樣本時(shí)，該樣本的K個(gè)鄰居中大容量類的樣本占多數(shù)。KKNNPythonkNNAABB??Python首先，我們新建一個(gè)kNN.py文件，文件里面包含兩個(gè)函數(shù)，一個(gè)用來生成小數(shù)據(jù)集，kNN分類算法。代碼如下：#kNN:kNearest#輸入 向#輸出fromnumpyimportimport#24def#group=array([[1.0,0.9],[1.0,1.0],[0.1,0.2],[0.0,4labels=['A','A','B',returngroup,KNNdefkNNClassify(newInput,dataSet,labels,numSamples #step1:tile(Areps):Areps#thefollowingcopynumSamplesrowsfordifftile(newInputnumSamples,1) #squaredDiffdiff** squaredDistsum(squaredDiffaxis #distancesquaredDist** step2:argsort()sortedDistIndices=classCount={}#defineadictionary(canbeappendelement)foriinxrange(k):step3:k abel=step4:k#whenthekeyvo abelisnotindictionaryclassCount,get()#willreturn0 abel]= abel,0)+##step5:返回出現(xiàn)次數(shù)最多的類 maxCount=forkey,valueinclassCount.items():ifvalue>maxCount:maxCount=valuemaxIndex=keyreturn 試importimportfromnumpyimport*dataSetlabelskNN.createDataSet()testX=array([1.2,1.0])k=3outputLabel=kNN.kNNClassify(testX,dataSet,labels,print"Yourinputis:",testX,"andclassifiedtoclass:",testX=array([0.1,outputLabel=kNN.kNNClassify(testX,dataSet,labels,print"Yourinputis:",testX,"andclassifiedtoclass:",YourYourinputis:[ 1.0]andclassifiedto Yourinputis:[ 0.3]andclassifiedto kNNknn算法來實(shí)現(xiàn)對手寫數(shù)字的自動(dòng)識別；0-9200每個(gè)樣本保持在一個(gè)txt32x32txt個(gè)數(shù)字，01 testDigits900KNN算法來實(shí)現(xiàn)。沒錯(cuò)，pythonkNN##kNN:kNearest#參數(shù)inX:vectortocomparetoexistingdatasetdataSet:sizemdatasetofknownvectors(NxM)labels:datasetlabels(1xMvector)k:numberofneighborstousefor#輸出fromnumpyimportimportos#KNN分 方defkNNClassify(newInput,dataSet,labels,numSamples #step1:tile(Areps):Areps#thefollowingcopynumSamplesrowsfordiff=tile(newInput,(numSamples,1))-dataSet #Subtractelement-wisesquaredDiff=diff**2#squaredforthesubtractsquaredDist=sum(squaredDiff,axis=1) #sumisperformedbyrowdistance=squaredDist**0.5step2:#sortedDistIndices=classCount #foriinstep3:k abel=step4:#whenthekeyvo abelisnotindictionaryclassCount,get()#willreturn0 abel]= abel,0)+step5:返回出現(xiàn)次數(shù)最多的類別作為分類結(jié)果maxCount=0forkey,valueinclassCount.items():ifvalue>maxCount:maxCount=valuemaxIndex=keyreturn# 轉(zhuǎn)換為向 rows=32cols=imgVector=zeros((1,rows*cols))fileIn=open(filename)forrowinlineStr=fileIn.readline()forcolinxrange(cols):imgVector[0,row*32+col]=int(lineStr[col])returnimgVector#def##step1: printGettingtrainingset..."dataSetDir==+' #numSamples=train_x=zeros((numSamples,1024))train_y=[]foriinfilename=#gettrain_x[i,:]=img2vector(dataSetDir+'trainingDigits/%s'%#getlabelfromfilenamesuchas"1_18.txt"label=int(filename.split('_')[0])#return1##step2: printGettingtestingset..."testingFileList=os.listdir(dataSetDir+'testDigits')#loadthetestingsetnumSamples=len(testingFileList)test_x=zeros((numSamples,1024))test_y=[]foriinfilename=#gettest_x[i,:]=img2vector(dataSetDir+'testDigits/%s'%#getlabelfromfilenamesuchas"1_18.txt"label=int(filename.split('_')[0])#return1###step1:加載數(shù)據(jù)print"step1:loadtrain_x,train_y,test_x,test_y=##step2:模型訓(xùn)練.printstep2training..."step3:print"step3:numTestSamples=test_x.shape[0]matchCount=0foriinpredict=kNNClassify(test_x[i],train_x,train_y,3)ifpredict==test_y[i]:matchCountmatchCount+=accuracy=float(matchCount)/step4:print"step4:showtheprint'Theclassifyaccuracyis:%.2f%%'%(accuracy*importimportstepstep1:loa