How can cluster users?

Hi,

I have a matrix where the first column is the user's id, I have 5 columns that represent the presence o absence of a variable (0 or 1 value). The matrix is:

    17     0     0     0     0     0
    18     0     0     0     0     0
    19     0     0     0     0     0
    20     1     0     0     1     0
    21     0     0     0     1     0
    22     1     1     1     1     1
    23     0     1     0     0     0

I want a technique, as classification or clusterization, to group users with the same values, for example: a class with all the users where all variables are not present, a class where all variables are present, etc. How can I perform this in Matlab? thanks.

Answer 1

[Walter Roberson]

binvec = mat2cell( dec2bin(0:31, 5) - '0', ones(32,1), 5);

binary_form = YourMatrix(:,2:end) * [16; 8; 4; 2; 1];  %note: algebraic matrix multiplication 
grouped = accumarray( binary_form(:) + 1, YourMatrix(:,1), [], @(L.') {L} );

group_table = [binvec, grouped];

Now, group_table will be a 32 x 2 cell array. group_table{K,1} will be a combination of variables, and group_table{K,2} will be the ID numbers that belong to that group. Note that group_table{K,2} might be empty.

Answer 2

[giuseppe]

Hi Walter,

thanks for your reply..but I have some questions. What is L? In the expression of grouped why there is @? Matlab don't recognizes this symbol and it gives an error. On L I received the error:

Undefined function or variable 'L'.

Error in bayesian (line 280) grouped = accumarray( binary_form(:) + 1, Z(:,1), [], (L.'), {L} );

Z is my matrix. thanks.

[Walter Roberson]

Sorry made a typo, should be

grouped = accumarray( binary_form(:) + 1, YourMatrix(:,1), [], @(L) {L.'} );

The expression

@(L) {L.'}

is equivalent to "handle of function T343203" where

function result = T343203(L)
  result = {L.'};
end

The syntax @(x) expression_involving_x is an "anonymous function"

[giuseppe]

Hi Walter, another question: how can I access to the elements of the cell array? If I try:

group_table{1}

ans =

     0     0     0     0     0

how can I find the id of grouped users? Thanks.