fcm.asv

FCM算法的matlab实现

function [Umat,Cmat,it,valJ] = fcm(Xdat,C,mfuz,epsilon,itmax);
%   function [Umat,Cmat,it,valJ] = fcm(Xdat,K,mfuz,epsilon,itmax);
%   模糊聚类主函数
%   
%   Xdat ：原数据集，大小为(M,N)
%   C    ：聚类数目
%   M：数据的个数，N：数据的维数
%   mfuz ：加权指数m
%   Umat ：模糊划分矩阵，大小为(C,M)
%   Cmat ：聚类中心矩阵，大小为(C,N)
%   Dmat ：距离矩阵，大小为(C,M)   
%   epsilon ：运算中止条件
%   itmax   ：最大迭代次数
%   it      ：迭代次数


load iris.txt;
Xdat = iris(:,3:6);
C=3;
mfuz=2;
epsilon=0.001;
itmax=100;
[M,N] = size(Xdat);
%   initialization of Cmat using random values
Cmat = fcm_dataInitC(Xdat,C);
Ct=Cmat;
it=1;
stab = 0;
while (it<itmax & stab ~=1)
%   compute matrix of distance
Dmat = fcm_calcD(Xdat,Cmat);
%   compute initial Umat
Umat = fcm_calcU(mfuz,Dmat);
Cmat = fcm_calcC(Xdat,C,mfuz,Umat);
Ct=Ct-Cmat;

%Ut = Umat;
%it=1; stab = 0;
%while (it<itmax & stab ~=1)
    %   compute Cmat, the matrix of centroids, from Xdat and Ut
 %   Cmat = fcm_calcC(Xdat,K,mfuz,Ut);
    %   Compute Dmat, the matrix of distances (euclidian)
  %  Dmat = fcm_calcD(Xdat,Cmat);
    %   update Umat, the fuzzy partition matrix
   % Umat = fcm_calcU(mfuz,Dmat);
    %   Calculate difference between Ut and Umat and put it in Ut
    %Ut = Ut-Umat;
    %fnorm = norm(Ut,'fro');
    %if (fnorm<epsilon)    stab = 1;   else    Ut = Umat;    end
    %it = it+1;
end


%valJ = fcm_evalJ(Umat,mfuz,Dmat);
return %% end of function fcm.m