function [predictions,erro] = predictor(classifier,projected_features,label_set,test_data,test_label_set,reductor_type,preserved)

disp('predictor')
size(test_data)
size(label_set)
dim=size(projected_features,1)

if strcmp(classifier,'fld')
    pf.X=projected_features; %tem de ser dim x num_data
    pf.y=label_set;  %1 x num_data
    modelfld = fld(pf);
    predictions = linclass(test_data,modelfld);
    disp('size prediction fld');
    size(predictions);
    %figure; ppatterns(pf); pline(modelfld);
    erro=cerror(predictions,test_label_set);
    %{
    test_struct.X = test_data;
    test_struct.y = test_label_set;
    figure; ppatterns(test_struct);
    title(strcat('Fisher LDA',', ,',reductor_type,', ,Dim: ',num2str(dim),', ,Error: ',num2str(erro),', ,InfoPreserved: ',num2str(preserved)));
    legend('1','2');
    xlabel('Feature 1') % x-axis label
    ylabel('Feature 2') % y-axis label
    zlabel('Feature 3') % <-axis label
    if size(projected_features,1)>2
       %plane3 %para desenhar plano
        plane3(modelfld);
    else

        pline(modelfld);
    end
    %}

end

%testar com minumum mahal
if strcmp(classifier,'minimum_mahal')

    predictions=zeros(size(test_data(1,:)));
    %mean1=mean(projected_features(label_set==1));
    %mean2=mean(projected_features(label_set==2));
    %cov1=cov(projected_features(label_set==1));
    %cov2=cov(projected_features(label_set==2));
    mean1=mean(projected_features(:,label_set==1),2);
    mean2=mean(projected_features(:,label_set==2),2);
    cov1=cov(projected_features(:,label_set==1)');
    cov2=cov(projected_features(:,label_set==2)');
    cov1=cov1';
    cov2=cov2';
    C=(cov1+cov2)/2;
    i2=(mean1-mean2)'/C;
    %SIMPLIFICAMOS PORQUE O MATLAB DIZIA QUE ASSIM ERA MAIS RAPIDO
    %inversaC = inv(C);
    %dif_means = mean1-mean2;
    %sum_means = mean1+mean2;
    %d = dif_means' * inversaC * ( test_data(:,i)' - 0.5 * sum_means);

    for i=1:size(test_data,2)
        d=i2*(test_data(:,i)-0.5*(mean1+mean2));

        if d>0
            predictions(i)=1;
        else
            predictions(i)=2;
        end
    end

    test_struct.X = test_data;
    test_struct.y = predictions;
    figure; ppatterns(test_struct);
    erro=cerror(predictions,test_label_set)
    title(strcat('Mahalanobis',', ,',reductor_type,', ,Dim: ',num2str(dim),', ,Error: ',num2str(erro),', ,InfoPreserved: ',num2str(preserved)));
    legend('1','2');
    xlabel('Feature 1') % x-axis label
    ylabel('Feature 2') % y-axis label
    zlabel('Feature 3') % <-axis label

    %desenhar original
    test_struct.X = test_data;
    test_struct.y = test_label_set;
    figure; ppatterns(test_struct);
    title(strcat('Original, ,Dim: ',num2str(dim)));
    legend('1','2');
    xlabel('Feature 1') % x-axis label
    ylabel('Feature 2') % y-axis label
    zlabel('Feature 3') % <-axis label


end
%projected_features, label_set, test_data
%é 1 e 2
%testar com minumum thigny         NAO ESTA A DAR
if strcmp(classifier,'minimum_eucl')
    predictions=zeros(size(test_data(1,:)));
    %mean1=mean(projected_features(label_set==1));
    %mean2=mean(projected_features(label_set==2));
    mean1=mean(projected_features(:,label_set==1),2);
    mean2=mean(projected_features(:,label_set==2),2);
    for i=1:size(test_data,2)
        d=(mean1-mean2)'*(test_data(:,i)-0.5*(mean1+mean2));
        if d>0
            predictions(i)=1;
        else
            predictions(i)=2;
        end
    end

    test_struct.X = test_data;
    test_struct.y = predictions;
    figure; ppatterns(test_struct);
    erro=cerror(predictions,test_label_set)
    title(strcat('Euclidian',', ,',reductor_type,', ,Dim: ',num2str(dim),', ,Error: ',num2str(erro),', ,InfoPreserved: ',num2str(preserved)));
    legend('1','2');
    xlabel('Feature 1') % x-axis label
    ylabel('Feature 2') % y-axis label
    zlabel('Feature 3') % <-axis label

    %desenhar original
    test_struct.X = test_data;
    test_struct.y = test_label_set;
    figure; ppatterns(test_struct);
    title(strcat('Original, ,Dim: ',num2str(dim)));
    legend('1','2');
    xlabel('Feature 1') % x-axis label
    ylabel('Feature 2') % y-axis label
    zlabel('Feature 3') % <-axis label

    %nao tenho model, nao consegui representar plano



end