Re: High utility item set mining
Date: August 16, 2020 10:34PM
Hi!
Interesting question. I think it depends on the application. For some applications, it makes sense to use the length constraint.
For example, if we analyze customer transactions for the purpose of marketing, maybe we don't want to find itemsets containing five or more items because it would be difficult to promote all these items together. For example, if we find that {juice, milk, bread, chocolate, noodle} yield a lot of money, it would not be very convenient to make some advertisement to buy five items together. Maybe it is easier to promote only 2 or 3 items together. So in such scenarios, I think it is fine to restrict the length of high utility itemsets.
Besides that, having the length constraint will not only eliminate very long itemsets but it will also make the discovery of itemsets faster. So it is easier to analyze bigger databases, and we can find the result more quickly. That is another advantage.
So I think it depends on the application, whether it is appropriate or not.
Best regards,