Majority filter（MF）

%%%%%测试MF和CF % load('C:\\Users\\cheng\\Documents\\Matlab\\UCI数据集\\sonar.mat'); % german_data=sonar(:,1:60); % german_label=sonar(:,61); % load('C:\\Users\\cheng\\Documents\\Matlab\\UCI数据集\\hearts.mat'); % german_data=hearts(:,1:12); % german_label=hearts(:,13); load('C:\\Users\\cheng\\Documents\\Matlab\\UCI数据集\\cleve.mat'); german_data=cleve(:,1:13); german_label=cleve(:,14); % load('C:\\Users\\cheng\\Documents\\Matlab\\UCI数据集\\austra.mat'); % german_data=austra(:,1:14); % german_label=austra(:,15); % load('C:\\Users\\cheng\\Documents\\Matlab\\UCI数据集\\breast.mat'); % german_data=breast(:,1:9); % german_label=breast(:,10); german_label=label_switch(german_label); data_set=normalize(german_data); data_label=german_label; [traindata,trainlabel,testdata,testlabel]=set_devide(data_set,data_label,4); [num_test,num_feature]=size(testdata); num_train=size(traindata,1); rate=0.1; [noisy_train_label,index]= add_noise(trainlabel, rate); num_noise=size(index,1); noisy_data=zeros(num_noise,num_feature); for i=1:num_noise noisy_data(i,:)=traindata(index(i),:); end [train_data,train_label,traindata1,trainlabel1]=set_devide(traindata,noisy_train_label,3); [traindata3,trainlabel3,traindata2,trainlabel2]=set_devide(train_data,train_label,2); %%%%%找traindata1中的噪声 train_data1=[traindata2;traindata3]; train_label1=[trainlabel2;trainlabel3]; % [~,belief1,~]=enn(train_data1,train_label1,5); % indices1=find(belief1>0.2); % num_indices1=size(indices1,1); % clear_data=zeros(num_indices1,num_feature); % clear_label=zeros(num_indices1,1); % for i=1:num_indices1 % clear_data(i,:)=train_data1(indices1(i),:); % clear_label(i,:)=train_label1(indices1(i),:); % end % train_data1=clear_data; % train_label1=clear_label; [ps,label1_tree]=BinaryClassify('tree',train_data1,train_label1,... traindata1,trainlabel1); [pn,label1_svm]=BinaryClassify('svm',train_data1,train_label1,... traindata1,trainlabel1); [pk,label1_3nn]=BinaryClassify('knn-3',train_data1,train_label1,... traindata1,trainlabel1); label1=[label1_tree,label1_svm,label1_3nn]; [index_noise1,~]=majority_filter(label1,trainlabel1); index1=find(index_noise1==1); num_index1=size(index1,1); potential_noise1=zeros(num_index1,num_feature); for i=1:num_index1 potential_noise1(i,:)=traindata1(index1(i),:); end % result_label1=vote(label1); % error_num_1=size(find((result_label1-testlabel)~=0),1); % en_grab_precision=1-error_num_1/num_test; label_bag=zeros(size(trainlabel1,1),15); for i=1:5 [new_traindata, new_trainlabel] =bootstrap(train_data1, train_label1); [~,label4_tree]=BinaryClassify('tree',new_traindata, new_trainlabel,... traindata1,trainlabel1); [~,label4_svm]=BinaryClassify('svm',new_traindata, new_trainlabel,... traindata1,trainlabel1); [~,label4_3nn]=BinaryClassify('knn-3',new_traindata, new_trainlabel,... traindata1,trainlabel1); label_bag(:,3*(i-1)+1)= label4_tree; label_bag(:,3*(i-1)+2)= label4_svm; label_bag(:,3*(i-1)+3)= label4_3nn; end index_noise4=majority_filter(label_bag,trainlabel1); index4=find(index_noise4==1); num_index4=size(index4,1); potential_noise4=zeros(num_index4,num_feature); for i=1:num_index4 potential_noise4(i,:)=traindata1(index4(i),:); end %%%%%找traindata2中的噪声 train_data2=[traindata1;traindata3]; train_label2=[trainlabel1;trainlabel3]; [~,label2_tree]=BinaryClassify('tree',train_data2,train_label2,... traindata2,trainlabel2); [~,label2_svm]=BinaryClassify('svm',train_data2,train_label2,... traindata2,trainlabel2); [~,label2_3nn]=BinaryClassify('knn-3',train_data2,train_label2,... traindata2,trainlabel2); label2=[label2_tree,label2_svm,label2_3nn]; index_noise2=majority_filter(label2,trainlabel2); index2=find(index_noise2==1); num_index2=size(index2,1); potential_noise2=zeros(num_index2,num_feature); for i=1:num_index2 potential_noise2(i,:)=traindata2(index2(i),:); end label_bag=zeros(size(trainlabel2,1),15); for i=1:5 [new_traindata, new_trainlabel] =bootstrap(train_data2, train_label2); [~,label5_tree]=BinaryClassify('tree',new_traindata, new_trainlabel,... traindata2,trainlabel2); [~,label5_svm]=BinaryClassify('svm',new_traindata, new_trainlabel,... traindata2,trainlabel2); [~,label5_3nn]=BinaryClassify('knn-3',new_traindata, new_trainlabel,... traindata2,trainlabel2); label_bag(:,3*(i-1)+1)= label5_tree; label_bag(:,3*(i-1)+2)= label5_svm; label_bag(:,3*(i-1)+3)= label5_3nn; end index_noise5=majority_filter(label_bag,trainlabel2); index5=find(index_noise5==1); num_index5=size(index5,1); potential_noise5=zeros(num_index5,num_feature); for i=1:num_index5 potential_noise5(i,:)=traindata2(index5(i),:); end %%%%%找traindata3中的噪声 train_data3=[traindata1;traindata2]; train_label3=[trainlabel1;trainlabel2]; [~,label3_tree]=BinaryClassify('tree',train_data3,train_label3,... traindata3,trainlabel3); [~,label3_svm]=BinaryClassify('svm',train_data3,train_label3,... traindata3,trainlabel3); [~,label3_3nn]=BinaryClassify('knn-3',train_data3,train_label3,... traindata3,trainlabel3); label3=[label3_tree,label3_svm,label3_3nn]; index_noise3=majority_filter(label3,trainlabel3); index3=find(index_noise3==1); num_index3=size(index3,1); potential_noise3=zeros(num_index3,num_feature); for i=1:num_index3 potential_noise3(i,:)=traindata3(index3(i),:); end label_bag=zeros(size(trainlabel3,1),15); for i=1:5 [new_traindata, new_trainlabel] =bootstrap(train_data3, train_label3); [~,label6_tree]=BinaryClassify('tree',new_traindata, new_trainlabel,... traindata3,trainlabel3); [~,label6_svm]=BinaryClassify('svm',new_traindata, new_trainlabel,... traindata3,trainlabel3); [~,label6_3nn]=BinaryClassify('knn-3',new_traindata, new_trainlabel,... traindata3,trainlabel3); label_bag(:,3*(i-1)+1)= label6_tree; label_bag(:,3*(i-1)+2)= label6_svm; label_bag(:,3*(i-1)+3)= label6_3nn; end index_noise6=majority_filter(label_bag,trainlabel3); index6=find(index_noise6==1); num_index6=size(index6,1); potential_noise6=zeros(num_index6,num_feature); for i=1:num_index6 potential_noise6(i,:)=traindata3(index6(i),:); end potential_noise_1=[potential_noise1;potential_noise2;potential_noise3]; potential_noise_2=[potential_noise4;potential_noise5;potential_noise6]; num1=find_same(noisy_data,potential_noise_1); num_1=num_index1+num_index2+num_index3; num2=find_same(noisy_data,potential_noise_2); num_2=num_index4+num_index5+num_index6; % for i=1: