最近邻法和k-近邻法学号:02105120 姓名::最近邻法:对于未知样本x,比较x与N个已知类别的样本之间的欧式距离,并决策x与距离它最近的样本同类。K近邻法:取未知样本x的k个近邻,看这k个近邻中多数属于哪一类,就把x归为哪一类。K取奇数,为了是避免k1=k2的情况。:要判别x属于哪一类,关键要求得与x最近的k个样本(当k=1时,即是最近邻法),然后判别这k个样本的多数属于哪一类。可采用欧式距离公式求得两个样本间的距离s=sqrt((x1-x2)^2+(y1-y2)^2):该算法中任取每类样本的一半作为训练样本,其余作为测试样本。例如iris中取每类样本的25组作为训练样本,剩余25组作为测试样本,依次求得与一测试样本x距离最近的k个样本,并判断k个样本多数属于哪一类,则x就属于哪类。测试10次,取10次分类正确率的平均值来检验算法的性能。:最近邻算实现对Iris分类clc;totalsum=0;forii=1:10data=load('');data1=data(1:50,1:4);%任取Iris-setosa数据的25组rbow1=randperm(50);trainsample1=data1(rbow1(:,1:25),1:4);rbow1(:,26:50)=sort(rbow1(:,26:50));%剩余的25组按行下标大小顺序排列testsample1=data1(rbow1(:,26:50),1:4);data2=data(51:100,1:4);%任取Iris-versicolor数据的25组rbow2=randperm(50);trainsample2=data2(rbow2(:,1:25),1:4);rbow2(:,26:50)=sort(rbow2(:,26:50));testsample2=data2(rbow2(:,26:50),1:4);data3=data(101:150,1:4);%任取Iris-virginica数据的25组rbow3=randperm(50);trainsample3=data3(rbow3(:,1:25),1:4);rbow3(:,26:50)=sort(rbow3(:,26:50));testsample3=data3(rbow3(:,26:50),1:4);trainsample=cat(1,trainsample1,trainsample2,trainsample3);%包含75组数据的样本集testsample=cat(1,testsample1,testsample2,testsample3);newchar=zeros(1,75);sum=0;[i,j]=size(trainsample);%i=60,j=4[u,v]=size(testsample);%u=90,v=4forx=1:ufory=1:iresult=sqrt((testsample(x,1)-trainsample(y,1))^2+(testsample(x,2)-trainsample(y,2))^2+(testsample(x,3)-trainsample(y,3))^2+(testsample(x,4)-trainsample(y,4)
模式识别+最近邻法和k近邻法matlab实现 来自淘豆网m.daumloan.com转载请标明出处.
