CAP_Toolbox/ComputeClusteringQuality.m
bb54cdf11825master

ComputeClusteringQuality.m
View Options

			function [CDF,Lorena] = ComputeClusteringQuality(Consensus,K_range)

			% Number of K values to check
			K = size(Consensus,3);

			% Number of pairs of frames
			n_items = size(Consensus,1);

			% Creates the CDF range
			c = 0:0.005:1;

			% Quality criterion computations are run for each explored K value...
			for k = 1:K

			% Sorted consensus entries
			Cons_val = sort(jUpperTriMatToVec(squeeze(Consensus(:,:,k))),'ascend');

			% Computation of CDF
			for i = 1:length(c)
			CDF(k,i) = sum(Cons_val <= c(i))/(n_items*(n_items-1)/2);
			end


			idx_dp = 1;

			for delta_perc = 0.5:0.5:10

			Lorena(k,idx_dp) = prctile(CDF(k,:),100-delta_perc) - prctile(CDF(k,:),delta_perc);
			idx_dp = idx_dp + 1;
			end




			% Computation of the AUC
			% AUC(k) = 0;
			%
			% for i = 2:(n_items*(n_items-1)/2)
			% AUC(k) = AUC(k) + (Cons_val(i)-Cons_val(i-1))* interp1q(c',CDF(k,:)',Cons_val(i));
			% end

			clear Cons_val
			end

			% for k = 2:K
			%
			% % Computation of Delta
			%
			% tmp_max_AUC = max(AUC(1:k-1));
			%
			% Delta(k) = (AUC(k) - tmp_max_AUC)/tmp_max_AUC;
			%
			% end
			%
			% Delta(1) = AUC(1);

			end