find repeating numbers

What is the best method to find the numbers in an array which repeat most frequently. For example, say that I have a matrix of values:

a = 1;
b = 30;
data = a + (b-a).*rand(1000,5);

I would like to find the five numbers which are repeated most in the matrix. They do not need to be in different columns/rows but simply the five most repeated numbers. How would I do this?

Answer 1

[Oleg Komarov]

unNum   = unique(data(:));
[n,bin] = histc(data(:),unNum);

With your example you're not likely to find any repetition (you should use randi).

The first output counts the repetitions while the second gives you which position in unNum is.

You can sort then n in decreasing order, get the sorting index and apply to unNum to retrieve the most repeated numbers.

An example:

a    = 0;
b    = 100;
data = randi([a b-a],1000,5);

unNum     = unique(data(:));
[n,bin]   = histc(data(:),unNum);

[srt,idx] = sort(n,'descend');
unNum(srt(1:5))

Answer 2

[Andrei Bobrov]

way with use accumarray

a    = 0;
b    = 100;
data = randi([a b-a],1000,5);
[unNum,n,n]     = unique(data(:));
c   = accumarray(n,1);
[srt,idx] = sort(c,'descend');
unNum(idx(1:5))

Answer 3

[dipanka tanu sarmah]

function y = repeated(x)
for i=1:length(x)
if length(find(x==i)) >=2
y{i,1}=x(i);
y{i,2}=length(find(x==i))
end
end
end

you can use this to find out the repeated values and frequency

[Stephen23]

This code is extremely unreliable, and should be avoided. In particular:

• This code does not find the values with the highest frequency, because it compares the array values with internally generated values (the loop index i). If any value inside x is not equal to one of 1:length(x) then this algorithm will not work. For example:

>> repeated([3,-2,-2]) % fails, no output
>> repeated([6,7,7])   % fails, no output

• The output of length changes depending on the sizes of the input array, so this code gives different solutions depending on the size of the input array x.
• length(find(...)) is slow and unnecessary. Simpler and faster would be to simply use nnz(...).
• The slow operations length(find(...)) are duplicated, whereas they should be simply called once.
• Bad code alignment is how beginners hide basic errors in their code. Code should be aligned using the MATLAB editor's default settings. Code can be aligned by simply selecting all code in the editor, then clicking ctrl+i.

Using histc (or the newer equivalent) is simpler and much more reliable than copying this buggy code.