"RapidMiner Association Rules / transaction based dataset"

TomLeey

Hello,

maybe someone here has already had experiences with this topic. I have itemsets that represent orders. The format  is:

OrderID    Item
1234        10
1234        12
1267        3
1288        3

My problem is, that the nominal-2-binominal operator doesn`t support this format. Using this Operator, it creates binominal datalines for each row. What i need is a grouping by the corresponding OrderID.

I spend a lot of time for finding a solution but i didn`t get any answers yet - although it seem to represent a common data-format.

Thanks for youre help !

Marco_Boeck

Hi,

I'm not sure what your desired output is. This?

OrderID    Item
1234        10, 12
1267        3
1288        3

Something else?

Regards,
Marco

TomLeey

Well, I think the FP-Growth Operator needs the following resulttable:

OrderID          Item_10      Item_12      Item_3   

1234              true            true            false
1267              false          false          true
1288              false          false          true

I want to search for association rules in the item sets. Therefore i have to use the FP-Growth Operator.

Thanks for your'e reply

Rene

check out this posting from haddock:
http://rapid-i.com/rapidforum/index.php/topic,1821.msg7227.html#msg7227

happy mining.

TomLeey

Is it also possible to do so without having an amount ? Surely its possible to use amount=1 but
the questions is whether there is a simpler solution.

Thanks for youre reply. I already had a look on the mentioned topic.