Idea for Hierarchical Clustering

%Idea for Hierarchical Clustering clc; clear all; close all; htree=[]; division=3; division=division+1; x_s=iris_dataset'; %------------------- load data % load spectra % x_s=NIR; %------------------- % load hald % x_s=hald; %------------------- % load fisheriris % x_sig=meas; %------------------- % load ionosphere % x_s=X'; %------------------- % load morse % x_s=Y0; %------------------- % load stockreturns % x_sig=stocks; x_s=x_s(:,:); count_clust=size(x_s,1); dend_array=1:size(x_s,1); %normalise brings each parameter of dataset in range of (0,1) normalise=[]; for i=1:size(x_s,2) x_main_s(:,i)=x_s(:,i)./max(x_s(:,i)); x_main=x_s(:,i)- x_main_s(:,i); end len=size(x_main_s,1); clus_count=3; x_main=x_main_s(1:len,:); position=zeros(1,len); N=1:len; x_main=[N' x_main]; x=x_main; dataset_len=size(x); mat=[]; %matrix formation for i=1:dataset_len(1,1)+1 for j=1:dataset_len(1,1)+1 C=0; if i==1 && j==1 mat(i,j)=0; elseif i==1 mat(i,j)=x(j-1,1); elseif j==1 mat(i,j)=x(i-1,1); else for b=1+1:dataset_len(1,2) C=C+(x(i-1,b)-x(j-1,b))*(x(i-1,b)-x(j-1,b)); end mat(i,j)=sqrt(C); end end end start=mat; z=[]; %clusterig the nearest elements in each loop for i=dataset_len(1,1):-1:division [x,mat,z,cr_values,small_index,p,count_clust,htree,dend_array]=cluster(x,mat,z,count_clust,htree,dend_array); end [htree]=plotd(x,mat,z,count_clust,htree,dend_array); H = dendrogram(htree,0); set(H,'LineWidth',1) title('Dendogram'); check=1; %using the returned list to give elements a cluster identity while check==1 check=0; for i=1:size(z,1) if(position(z(i,1))==0 && position(z(i,2))==0) position(z(i,2))=clus_count; position(z(i,1))=clus_count; clus_count=clus_count+1; elseif (position(z(i,1))==0 || position(z(i,2))==0 ) position(z(i,2))=max(position(z(i,1)),position(z(i,2))); position(z(i,1))=max(position(z(i,1)),position(z(i,2))); elseif (position(z(i,1))>=position(z(i,2)) ) position(z(i,1))=position(z(i,2)); else position(z(i,2))=position(z(i,1)); check=1; end end end position=[N ;position]'; %sorting cluster identity final_pos=sortrows(position,2); change=final_pos(1,2); start_end=[]; count=1; start_end(count,1)=1; last=0; %calculating start and end of each cluster for i=1:size(final_pos,1) if(final_pos(i,2)~=change) start_end(count,2)=i-1; change=final_pos(i,2); count=count+1; start_end(count,1)=i; end end start_end(count,2)=size(final_pos,1); grpa=([final_pos(start_end(1,1):start_end(1,2),1)]'); grpb=([final_pos(start_end(2,1):start_end(2,2),1)]'); grpc=([final_pos(start_end(3,1):start_end(3,2),1)]'); count_cluster=[length(grpa),length(grpb),length(grpc)]; color=['b' 'r' 'g' 'y' 'k']; symbol=['*']; x=1;y=2;z=3; figure; scatter3(x_s(grpa,x),x_s(grpa,y),x_s(grpa,z),symbol(1),color(1));hold on; scatter3(x_s(grpb,x),x_s(grpb,y),x_s(grpb,z),symbol(1),color(2));hold on; scatter3(x_s(grpc,x),x_s(grpc,y),x_s(grpc,z),symbol(1),color(3));hold on; title('Reduced set of prototype ' ); xlabel('Featuer x');ylabel('Featuer y'); grid on x=2;y=3;z=4; figure; scatter3(x_s(grpa,x),x_s(grpa,y),x_s(grpa,z),symbol(1),color(1));hold on; scatter3(x_s(grpb,x),x_s(grpb,y),x_s(grpb,z),symbol(1),color(2));hold on; scatter3(x_s(grpc,x),x_s(grpc,y),x_s(grpc,z),symbol(1),color(3));hold on; title('Reduced set of prototype ' ); xlabel('Featuer x');ylabel('Featuer y'); grid on x=3;y=4;z=1; figure; scatter3(x_s(grpa,x),x_s(grpa,y),x_s(grpa,z),symbol(1),color(1));hold on; scatter3(x_s(grpb,x),x_s(grpb,y),x_s(grpb,z),symbol(1),color(2));hold on; scatter3(x_s(grpc,x),x_s(grpc,y),x_s(grpc,z),symbol(1),color(3));hold on; title('Reduced set of prototype ' ); xlabel('Featuer x');ylabel('Featuer y'); grid on x=4;y=1;z=2; figure; scatter3(x_s(grpa,x),x_s(grpa,y),x_s(grpa,z),symbol(1),color(1));hold on; scatter3(x_s(grpb,x),x_s(grpb,y),x_s(grpb,z),symbol(1),color(2));hold on; scatter3(x_s(grpc,x),x_s(grpc,y),x_s(grpc,z),symbol(1),color(3));hold on; title('Reduced set of prototype ' ); xlabel('Featuer x');ylabel('Featuer y'); grid on