I'm reading about the Apriori algorithm using the textbook Introduction to Machine Learning (Ethem Alpaydin) and had a question.

I've noticed that the textbook and many other resources I find online say that the basic intuition behind Apriori is that in order for an itemset to be "frequent," its subsets should also be frequent.

What is the criteria for something to be considered frequent? It seems that we want to make sure that the itemset has a sufficient amount of support, but that also seems subjective.

Is this determined empirically?


A threshold for an interesting frequency of rules is subjective, and depends a fair bit on the purpose. Often you will pick a support threshold that leaves you with a manageable number of rules to achieve that purpose. However, support is not the only measure of 'interestingness' available, and using another measure as well as support may allow you to have a lower support threshold without having too many rules to deal with.

Check out this book, particularly chapter 5 (available as a sample chapter) for more info.


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.