Suppose i have matrix the first and second column represent x,y-coordinates and third column is cluster id

   95.0129    5.7891    3.0000
   23.1139   35.2868    1.0000
   60.6843   81.3166    2.0000
   48.5982    0.9861    3.0000
   89.1299   13.8891    3.0000
   76.2097   20.2765    3.0000
   45.6468   19.8722    3.0000
    1.8504   60.3792    1.0000
   82.1407   27.2188    3.0000
   44.4703   19.8814    3.0000
   61.5432    1.5274    3.0000
   79.1937   74.6786    2.0000
   92.1813   44.5096    2.0000
   73.8207   93.1815    2.0000
   17.6266   46.5994    1.0000
   40.5706   41.8649    1.0000
   93.5470   84.6221    2.0000
   91.6904   52.5152    2.0000
   41.0270   20.2647    3.0000
   89.3650   67.2137    2.0000

I want create the matrix of individual cluster id with its x,y coordinates, for example cluster 1; c(1)= [23.1139 35.2868; 1.8504 60.3792; 17.6266 46.5994; 40.5706 41.8649]

Answer 1

[Stephen23]

Method One: accumarray:

>> R = 1:size(M,1);
>> C = accumarray(M(:,3),R(:),[],@(r){M(r,1:2)});
>> C{1}
ans =
   23.1139   35.2868
    1.8504   60.3792
   17.6266   46.5994
   40.5706   41.8649
>> C{2}
ans =
   60.6843   81.3166
   79.1937   74.6786
   92.1813   44.5096
   73.8207   93.1815
   93.5470   84.6221
   91.6904   52.5152
   89.3650   67.2137
>> C{3}
ans =
   95.0129    5.7891
   48.5982    0.9861
   89.1299   13.8891
   76.2097   20.2765
   45.6468   19.8722
   82.1407   27.2188
   44.4703   19.8814
   61.5432    1.5274
   41.0270   20.2647

Method Two: arrayfun:

>> fun = @(r) M(M(:,3)==r,1:2);
>> C = arrayfun(fun, unique(M(:,3)), 'uni',false);
>> C{1}
ans =
   23.1139   35.2868
    1.8504   60.3792
   17.6266   46.5994
   40.5706   41.8649
>> C{2}
ans =
   60.6843   81.3166
   79.1937   74.6786
   92.1813   44.5096
   73.8207   93.1815
   93.5470   84.6221
   91.6904   52.5152
   89.3650   67.2137
>> C{3}
ans =
   95.0129    5.7891
   48.5982    0.9861
   89.1299   13.8891
   76.2097   20.2765
   45.6468   19.8722
   82.1407   27.2188
   44.4703   19.8814
   61.5432    1.5274
   41.0270   20.2647

[Juma Ally]

thank you

[Juma Ally]

Thank you for the answer; As i consider x,y-coordinates represent the location of the nodes in x,y-plan and i assume the nodes in each cluster are correlated based on overlapping. I want to compute the correlation coefficient by first compute the euclidean distance between the nodes and then using exponential model to compute correlation coefficient for example c(1,1) = exp(-d(1,1)) then i can get matrix for each cluster. Thank you