Can someone tell me which algorithm would be best to solve this problem? I am thinking this belongs in the category of 'Market Basket Analysis' aka ('Affinity Analysis'), but I am not sure.

Here's the problem:

A user belongs to certain types of groups. When our client gives us a set of users that belong to a particular group, we need to tell them which other groups those users might be most interested in as well as most disinterested in.

To me this sounds similar to the classic Market Basket Analysis example: 'customers who buy diapers also buy beer'. So I was thinking - if I pre-compute all the affinities within groups then I will be able to answer our client's question. No?

Would greatly appreciate any thoughts on this? What kind of algorithm would be most appropriate for this?

Thanks for your help.


Yes, this can be solved with Market Basket Analysis.

Since Market Basket Analysis is a special case of Recommender Systems, so you will find a lot of helpful algorithms in this area, too. This question seems to be similar to yours Algorithm to calculate difference in users' tastes, additionally you might find Recommendation for a book about recommender systems helpful.

However, if you do not have explicit information about particular user dislikes, you will have a hard time to distinguish "dislike" from "does not know about" i.e. "not sure whether likes or not". But I would not worry about this, the most important task is to recommend the most interesting groups ... the rest is simply not shown.

| cite | improve this answer | |

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.