聚類演算法Kmeans—–matlab code

function [cid,nr,centers] = cskmeans(x,k,nc)
% CSKMEANS K-Means clustering - general method.
%
%   This implements the more general k-means algorithm, where
%   HMEANS is used to find the initial partition and then each
%   observation is examined for further improvements in minimizing
%   the within-group sum of squares.
%
%   [CID,NR,CENTERS] = CSKMEANS(X,K,NC) Performs K-means
%   clustering using the data given in X.
%  
%   INPUTS: X is the n x d matrix of data,
%   where each row indicates an observation. K indicates
%   the number of desired clusters. NC is a k x d matrix for the
%   initial cluster centers. If NC is not specified, then the
%   centers will be randomly chosen from the observations.
%
%   OUTPUTS: CID provides a set of n indexes indicating cluster
%   membership for each point. NR is the number of observations
%   in each cluster. CENTERS is a matrix, where each row
%   corresponds to a cluster center.
%
%   See also CSHMEANS

%   W. L. and A. R. Martinez, 9/15/01
%   Computational Statistics Toolbox

warning off
[n,d] = size(x);
if nargin < 3
 % Then pick some observations to be the cluster centers.
    ind = ceil(n*rand(1,k));
 % We will add some noise to make it interesting.
 nc = x(ind,:) + randn(k,d);
end

% set up storage
% integer 1,...,k indicating cluster membership
cid = zeros(1,n); 
% Make this different to get the loop started.
oldcid = ones(1,n);
% The number in each cluster.
nr = zeros(1,k); 
% Set up maximum number of iterations.
maxiter = 100;
iter = 1;

while ~isequal(cid,oldcid) & iter < maxiter
 % Implement the hmeans algorithm
 % For each point, find the distance to all cluster centers
 for i = 1:n
  dist = sum((repmat(x(i,:),k,1)-nc).^2,2);
  [m,ind] = min(dist); % assign it to this cluster center
  cid(i) = ind;
 end
 % Find the new cluster centers
 for i = 1:k
  % find all points in this cluster
  ind = find(cid==i);
  % find the centroid
  nc(i,:) = mean(x(ind,:));
  % Find the number in each cluster;
  nr(i) = length(ind);
 end
  iter = iter + 1;
end

% Now check each observation to see if the error can be minimized some more.
% Loop through all points.
maxiter = 2;
iter = 1;
move = 1;
while iter < maxiter & move ~= 0
 move = 0;
 % Loop through all points.
 for i = 1:n
  % find the distance to all cluster centers
  dist = sum((repmat(x(i,:),k,1)-nc).^2,2);
  r = cid(i);  % This is the cluster id for x
        %%nr,nr+1;
  dadj = nr./(nr+1).*dist'; % All adjusted distances
  [m,ind] = min(dadj); % minimum should be the cluster it belongs to
  if ind ~= r  % if not, then move x
   cid(i) = ind;
   ic = find(cid == ind);
   nc(ind,:) = mean(x(ic,:));
   move = 1;
  end
 end
 iter = iter+1;
end
centers = nc;
if move == 0
 disp('No points were moved after the initial clustering procedure.')
else
 disp('Some points were moved after the initial clustering procedure.')
end
warning on