How to run k means clustering on multiple items in a text file

I am trying to find a way to run k-means clustering on data from the attached text file. The first column shows an individual item. for example: A8888888888888888880011 is one item and it appears several times in the data. I would like to run k means clustering for each item in the text file (the one provided is shortened by a few thousand lines). I read the data in and sort it by using:

fid = fopen(filename);

cell_data = textscan(fid, '%s %n %n');

fclose(fid);

matrix_data = [cell_data{:}];

sort_data = sortrows(matrix_data, [1,3]);

After this I would like to have one coordinate for each item. The coordinate being the result of k means clustering returning a point (column b, column c).

So after running this on the text file, I would end up with 3 points, one for item 11, 12, and 13.

I hope this makes sense but if there are any questions please ask

[sixwwwwww]

Dear Aaron, Can you please explain what k-mean clustering is?

[Aaron]

I will just give you an example of what I am wanting to happen because I am not entirely sure on the function to use. Lets say I have a plot of points such as the following:

....................................

                                 x
  x x
x  y  x
 x  x x            x                     x

I would like it to ignore the 3 outliers and return a point that is somewhere near y. The best way I know to explain what I am wanting is that wherever the biggest cluster of data is, I want one point that best represents that cluster. Do you know the best way to go about this in matlab?

[sixwwwwww]

In your example, you have a cluster of 8 elements(7 x and 1 y). So can we take mean of 8 points as a cluster point? or should we select 1 of these 8 points strictly?

[Jonathan LeSage]

Could you clarify what data you are trying to cluster in your example? Are you trying to cluster column one data or column two as one-dimensional vectors? Or are you trying to cluster two-dimensional data?

Answer 1

[Yatin]

Hi,

You can try using the "kmeans" function by formatting your data as per the data format required in the function. The function returns the centroid of the cluster which is the best representation of the cluster. For more information you can refer to the link below:

http://www.mathworks.com/help/stats/kmeans.html