Index number of k-means clusters

Hi guys, I have a program that cluster some data and then calculate the optimized number of clusters.

I want to know how I can get the index number of each data so I can plot them separated in clusters

My code is the following:

%importar dados do excel
imp = xlsread('Academia.xlsx');

%%Loop Kmeans K clusters

k=20

CH=zeros(1,k);
SH=zeros(1,k);
DB=zeros(1,k);
SUB=zeros(1,k);

% 
% for i=1:k
%     CH{i}=0;
%     SH{i}=0;
%     DB{i}=0;
% end

% eva = evalclusters(x,clust,criterion)
for i=1:k
    [idx,C]=kmeans(imp,i,'MaxIter',10000); %C=centroides
%  Como pegar o valor de idx ?
    eva = evalclusters(imp, idx, 'CalinskiHarabasz')
    CH(1,i)=eva.CriterionValues
    eva2 = evalclusters(imp, idx, 'Silhouette')
    SH(1,i)=eva2.CriterionValues
    eva3 = evalclusters(imp, idx, 'DaviesBouldin')
    DB(1,i)=eva3.CriterionValues
end

CH(1,1)=0
% %  Descobrir as diferenças entre os valores de CH
while i>1
    SUB(1,i)=CH(1,i)-CH(1,i-1);
    i=i-1;
end

SUB(1,1)=0
%Achar o pulo maximo entre os clusters para o valor de CH:
%  cell2mat(SUB); %Converte para matriz
%SUB2=cell2mat(SUB)

   [V,N]=max(SUB) % Por algum motivo está pulando a célula vazia e fornecendo valor incorreto

   SH
   %SHF=cell2mat(SH)
   [V2,N2]=max(SH) %Valor mais próximo de 1

   DB
   %DBF=cell2mat(DB)
   [V3,N3]=min(DB) %Valor mais próximo de 0

if (N==N2) && (N2==N3)
    disp('CH=SH=DB')
    N
    N2
    N3
    i=N
%     y=xlswrite('Academia_target.xlsx',M(1,i),'D1:D80')
elseif N2==N3  
    disp('SH=DB')
elseif N==N3
    disp('CH=DB')
elseif N==N2
    disp('CH=SH')
else
    disp('Todas as métricas forneceram valores diferentes')
end

In this case I want to retrieve the idx from the clustering in 4 clusters which is the optimized on in my case ( the Excel file from which I take the data is attached to this question.

Thanks in Advance !

[jgg]

I'm unclear what the problem is here; idx is a vector corresponding to the cluster ID for each observation for your fitted cluster. Isn't that what you want?

[Gabriel]

Hi jgg,

The thing is that each time the K-means runs (up to 20) the value of idx changes. And I only know which value "i" I need after the loop for is completed. So I need to store all of idx values and have a way to retrieve the value which corresponds to i=4 in this case.

I already tried using a variable R{i}=idx after the k-means but for some reason I receive a error message saying (Cell contents assignment to a non-cell array object.

Error in import_excelv4 (line 27) R{i}=idx

Answer 1

[wathiq dukhan]

I would like a program to calculate the number of clusters by genetic algorithm and k-means.

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

k-means itself must be passed the number of clusters to use; it is not able to calculate the number of clusters. However, there are algorithms that can be used that run k-means a number of times and take estimates of what the most likely number of clusters is under certain conditions.

[wathiq dukhan]

I need to use GA to calculate the number of clusters.