When preparing my data for funneling into the Microsoft Association Rules algorithm, I was not sure if I should group by data by Transaction and Item, or have a record for every instance of an item in a transaction. Does the algorithm care and add weight if an item appears 3 times in a transaction Or is it just looking for the existence of an item with another item, regardless of how many are present?
1 Answer
No it is not important and highly recommended to remove the duplicate items and sort the items in lexicographical order in each transaction. This is to improve the performance.
In association rule mining, an item is frequent iff it is repeated in multiple transactions not in a single transaction. This is why you don't need to have duplicate items in a each transaction.