Let's say we have a sport, where many competitions are organized throughout the year. In these events, a set of people compete, although not every person competes in every event.

I would like to categorize these people, and the events, in such a way that people in category $A$ performed well in the events of category $A$ etc. This also means that people in category $A$ usually perform well at the same events. There would then be a one-to-one correspondence between event categories and people categories.

It would also be OK for me if every event and every person is not uniquely put into one category, but is given a weight $w_A$ representing how much it fits into category $A$

The data I have are the results from all the events.

What do you think would be a good algorithm to categorize both events and people at the same time (since they are related)? I have been reading up on different machine learning algorithms, but would love some input on what to focus on.

  • $\begingroup$ You should look into biclustering. From my understanding, it fits your description exactly, but I've never worked with the technique, so I can't provide much besides a search term. $\endgroup$ – alto Nov 5 '13 at 22:45
  • $\begingroup$ Thanks, I will look into that. The big question is how to deal with missing data. I have a lot of that since not every competitor participates in every event. $\endgroup$ – burk Nov 17 '13 at 20:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.