KNN.rar_K._K近邻算法_Nearest neighbour _knn_k近邻matlab实现

whos; iris_original=importdata('iris.data'); iris_testset=zeros(50:5); iris_trainingset=zeros(100:5); distance=zeros(50:1); class1=0; class2=0; class3=0; tmpnum=0; j=1; i=1; testcases=50; trainingcases=100; nbofmatch=0; nbofnomatch=0; matchpercentage=0; nomatchpercentage=0; matcharray=zeros(99,1); comfusion_matrix=zeros(3:3); nboftmp1=0; nboftmp2=0; nboftmp3=0; l=1; k=1; % Choose the training data from the original data for i=1:150 if (i <=34 | i > 50 & i <= 83 | i > 100 & i <= 133) iris_trainingset(j,1:5)=iris_original(i,1:5); j=j+1; elseif (i > 34 & i <= 50 | i > 83 & i <= 100 | i > 133 & i <= 150) iris_testset(l,1:5)=iris_original(i,1:5); l=l+1; end end % Choose the test data from the original data for k=1:99 for j=1:testcases for i=1:100 distance(i,1) = sqrt((iris_testset(j,1) - iris_trainingset(i,1))^2 + (iris_testset(j,2) - iris_trainingset(i,2))^2 + (iris_testset(j,3) - iris_trainingset(i,3))^2 + (iris_testset(j,4) - iris_trainingset(i,4))^2); distance(i,2) = iris_trainingset(i,5); end % Bobble sort distance with their label for m=1:100 for n=1:99 if (distance(n+1,1) < distance(n,1)) tmp=distance(n+1,1); label_tmp=distance(n+1,2); distance(n+1,1)=distance(n,1); distance(n+1,2)=distance(n,2); distance(n,1)=tmp; distance(n,2)=label_tmp; end end end % Calculate number of class1, class2, class3 for s=1:k if (distance(s,2) == 1) class1=class1+1; elseif (distance(s,2) == 2) class2=class2+1; elseif (distance(s,2) == 3) class3=class3+1; end end % calculate which class have the majority number and assign a label if (class1 > class2 & class1 > class3) label=1; elseif (class2 > class3 & class2 > class1) label=2; elseif (class3 > class2 & class3 > class1) label=3; elseif (class1 > class2 & class1 == class3) label=distance(1,2); elseif (class2 > class3 & class2 == class1) label=distance(1,2); elseif (class3 > class1 & class2 == class3) label=distance(1,2); elseif (class1 == class2 & class1 == class3) label=distance(1,2); end % calculate number of match label and number of nomatch label if (iris_testset(j,5) == label) nbofmatch=nbofmatch+1; else nbofnomatch=nbofnomatch+1; end class1=0; class2=0; class3=0; % comfusion_matrix if (label == 1) if (iris_testset(j,5) == 1) comfusion_matrix(1,1) = comfusion_matrix(1,1) + 1; elseif (iris_testset(j,5) == 2) comfusion_matrix(1,2) = comfusion_matrix(1,2) + 1; elseif (iris_testset(j,5) == 3) comfusion_matrix(1,3) = comfusion_matrix(1,3) + 1; end elseif (label == 2) if (iris_testset(j,5) == 1) comfusion_matrix(2,1) = comfusion_matrix(2,1) + 1; elseif (iris_testset(j,5) == 2) comfusion_matrix(2,2) = comfusion_matrix(2,2) + 1; elseif (iris_testset(j,5) == 3) comfusion_matrix(2,3) = comfusion_matrix(2,3) + 1; end elseif (label == 3) if (iris_testset(j,5) == 1) comfusion_matrix(3,1) = comfusion_matrix(3,1) + 1; elseif (iris_testset(j,5) == 2) comfusion_matrix(3,2) = comfusion_matrix(3,2) + 1; elseif (iris_testset(j,5) == 3) comfusion_matrix(3,3) = comfusion_matrix(3,3) + 1; end end end % end of j comfusion_matrix comfusion_matrix=zeros(3:3); matchpercentage = (nbofmatch/testcases)*100; matcharray(k,1)=matchpercentage; matchpercentage=0; nomatchpercentage=0; nbofmatch=0; nbofnomatch=0; end % end of k % Plot the graph x=linspace(1,99,99); title('iris for KNN'); xlabel('Number of K'); ylabel('Match Percentage %'); grid on; plot(x,matcharray);