Error node "subset matcher"

In the workflow “Market Basket Analysis: Deploy Association Rules”, the “Subset Matcher” node takes two collection columns as input: the antecedents in the rule set (top input port) and the content of the current shopping basket (lower input port). It matches the current basket item set with all possible subsets in the rule antecedent item sets. The output table contains pairs of matching cells: the current shopping basket and the total or partial matching antecedents from the rule set, but in the output table “AntecedentsItems” the values ​​are ordered and do not match the antecedents in the rule set. For example, if you enter the basket [229, 225], the output table “AntecedentsItems” is [225], [225,229] and [229]. The association [229, 225] is not found and no recommended products are located.

Hi jablanco81,

Would it be possible for you to share an example workflow?

Thank you,
Cheers,
Vincenzo

Hi Vicenzo,

I share example workflow. My shopping list is (229,225). The list (229,225) exists in the AssociationRules.table node and the recommended product is blueberry.

In the node “Subset Matcher” the output table “AntecedentsItems” is [225], [225,229] and [229]. The association [229, 225] is not found and no recommended products are located (blueberry).

Thank you,
José A.

Using Association Rules for Market Basket Analysis. List [229, 225].knwf (63.5 KB)

Hi José,

The workflow does not contain the data files. Would it be possible to share those?

Thank you,
Vincenzo

Hi Vicenzo,

the size of the file with the data is 3.7 MB and it is not possible to upload it on this platform.
How can I share it?

This is the same case as the “Using Association Rules for Market Basket Analysis” workflow by inserting list 229 225 (attached image) in the “Table Creator” node.

Thank you,
José A.

I guess this topic has been solved in this thread Subset Matcher

Hi Vicenzo,

If the order of appearance of the products must be taken into account in order to carry out the association rules, the recommendation engine is NOT correct, since the “subset matcher” node returns the ordered lists.
The node “subset matcher” must not order the lists it returns.

I have ordered the association rule lists so that the recommendation engine is correct but it is NOT the solution to the problem. The lists of the association rule do not have to be ordered. Order is important in many recommendation engines.

Thank you,
José A.

Hi José,

I just checked it out. You are right. The solution right now would be to order the items in the association rule set dataset.

I will discuss it with the development team and will come back to you asap.

Thank you,
Cheers,
Vincenzo

Hi Vicenzo,

I’m waiting for news.

Thank you,
Cheers,
José A.