mcd_gauss_dd.m

数据挖掘的工具箱,最新版的,希望对做这方面研究的人有用

%MCD_GAUSS_DD Minimum Covariance Determinant Robust Gaussian data description.
% 
%       W = MCD_GAUSS_DD(A,FRACREJ)
% 
% Fit a Minimum-Covariance-Determinant Gaussian density on dataset A. The
% algorithm is taken from :
%
%  Rousseeuw, P.J. and Van Driessen, Katrien, "A fast algorithm for
%  the minimum covariance determinant estimator", 15 Dec. 1998
% 
% See also: datasets, mappings, dd_roc, fastmcd, gauss_dd

% Copyright: D. Tax, R.P.W. Duin, [e-mail]
% Faculty of Applied Physics, Delft University of Technology
% P.O. Box 5046, 2600 GA Delft, The Netherlands
  
function [W,out] = mcd_gauss_dd(x,fracrej)

if nargin < 2 | isempty(fracrej), fracrej = 0.05; end
if nargin < 1 | isempty(x) 
	W = mapping(mfilename,{fracrej});
	W = setname(W,'Minimum Covariance Determinant Gaussian');
	return
end

if isa(fracrej,'double')           %training

	x = +target_class(x);     % only use the target class
	dim = size(x,2);

	% call the function:
	options.alpha = 1-fracrej;
	options.lts = []; % display nothing!
	res = fastmcd(x,options);

	% obtain the threshold:
	p = mahaldist(x,res.center,res.cov,-1);
	thr = dd_threshold(p,1-fracrej);

	%and save all useful data:
	W.m = res.center;
	W.s = res.cov;
	W.threshold = thr;
	%W = mapping(mfilename,'trained',W,str2mat('target','outlier'),dim,2);

	% actually, we can just call 'gauss_dd' now!
	W = mapping('gauss_dd','trained',W,str2mat('target','outlier'),dim,2);
	W = setname(W,'Minimum Covariance Determinant Gaussian');

else
	error('Evaluation of mcd_gauss_dd is treated by gauss_dd!');
end
return