close all; clear all; clc;                                                                         
%add dataset
     wholeData = csvread('');

for foldNO =1:10    
                                                                                  
 A = wholeData;
 A(A<0)=0;
 Data = A(:,1:end-1);
 labels  = A(:,end); 
 
 ind = crossvalind('Kfold',labels,10); 
 test= ind==foldNO;
 train = ~test;
 
 trainData = Data(train,:); trainLabels = labels(train); 
 testData = Data(test, :); testLabels = labels(test);

%     class0 = [];
%     class1 = [];
%     for ii = 1:size(A, 1)
%         if(A(ii,(size(A, 2)))==0)
%             class0 = [class0; A(ii,1:(size(A, 2)-1))];
%         else
%             class1 = [class1; A(ii,1:(size(A, 2)-1))];
%         end
%     end
    
    class0 = [];
    class1 = [];
    for ii = 1:size(trainData, 1)
        if(trainLabels(ii)==0)
            class0 = [class0; trainData(ii,:)];
        else
            class1 = [class1; trainData(ii,:)];
        end
    end
    
 train_data= {class0',class1'};
 train_labels=[1 2];
 
 test_data = testData';
 test_labels = testLabels' + 1;


    %warning off
    param.numHiddenLayers = 1;
    param.numHiddenNeurons = [3,3,10];
    CC = 0;
    param.C = [0.01,0,100];
    param.rhoValue = 0.0001;
    sg = 0.001;
    param.sigpara = [sg,sg,sg];
    param.sigpara1 = [sg,sg,sg];
    
    for m = 1:size(train_data,2)
        dt = normc(train_data{m});
        %dt = train_data{m};
        model{m} = ELM_AE(dt, param);
    end    
    test_data = normc(test_data);
    num_test_samples = size(test_data,2);
    
    predicted_labels=[];
    
    for jj=1:num_test_samples
        test_sample = test_data(:,jj);
        tmpErr = [];
        for ii =1:size(train_data,2)
            TY = ELM_AE_Reconstruct(test_sample, model{ii});
            tmpErr(ii) = sqrt(sum((TY-test_sample).^2));
        end
        [minVal,lb]=min(tmpErr,[],2);
        predicted_labels(jj) = lb(1);
    end
    acc = sum(test_labels==predicted_labels)/num_test_samples;
    fprintf('\naccuracy=%f\n', acc);
    accuracy(foldNO) = acc;

end
% mean(accuracy)