ide.m

The matlab code implements the ensemble of decision tree classifiers proposed in: "L. Nanni and A. L

%the code for treec and pca is in the PRTools 3.1.7 (it is a matlab free toolbox)
%the code for NPE is available at
%http://www.cs.uiuc.edu/homes/dengcai2/Data/data.html

function [SCORE]=IDE(TR,TE,y,yy, Nensemble)
%TR= training set with label y
%TE test set with label yy
% Nensemble is the number of classifiers build for each class

for i=1:max(y)
    for ii=1: Nensemble
        % bagging like selection
        [aa,bb]=sort(rand(size(TR(find(y==i),:),1),1));
        Ftr=TR(find(y==i),:);
        Ftr=Ftr(bb(1:size(Ftr,1)*0.637),:);
        %creation of the PCA space
        [W] = pca(dataset(Ftr),0.98);
        %projection on the PCA space
        NTR=+(W*dataset(TR));
        NTE=dataset(TE)*W;
        clear W
        options = [];
        options.k = 5;
        options.NeighborMode = 'Supervised';
        options.gnd = y;
        [eigvector, eigvalue] = NPE(options, NTR);
        NTR = NTR*eigvector;
        NTE = NTE*eigvector;
        w7=treec(dataset(NTR,y'),'infcrit',2);
        clear NTR
        if ii==1 & i==1
            SCORE=+(w7*dataset(NTE));
        else
            SCORE=SCORE+(w7*dataset(NTE));
        end
    end
end