How to test/validate unlabeled data in association rules in R?

Question

I produced association rules by using the arules package (apriori). I'm left with +/- 250 rules. I would like to test/validate the rules that I have, like answering the question: How do I know that these association rules are true? How can I validate them? What are common practice to test it?

I thought about cross validation (with training data and test data) as I read that it's not impossible to use it on unsupervised learning methods..but I'm not sure if it makes sense since I don't use labeled data.

If someone has a clue, even if it's not specifically about association rules (but testing other unsupervised learning methods), that would also be helpful to me.

I uploaded an example of the data that I use here in case it's relevant: https://www.mediafire.com/?4b1zqpkbjf15iuy

Answer 1

You may want to consider using your own APparameter object to put "significance" constraints on the rules learned by Apriori. See page 13 of the arules documentation. This could reduce the number of uninteresting rules returned in your run.

In lieu of gold standard data for your domain, consider bootstrap resampling as a form of validation, as described in this article.