Market basket analysis data preprocess

newjop228

Hi all! I am extremely new to Rapidminer (only got introduced recently). I am trying to conduct a market basket analysis with the given CSV file. The example is something like this:

ID,Item
C1,yogurt,cheese roll,cat pork
C2,chicken,yogurt,pork,soda,whipped/sour cream
C3,beef
C4,onions,liquor
C5,soda,whipped/sour cream
...
C2000,soda,cheese roll,yogurt

I understand I will need to apply FP-Growth followed by Association rule in Rapidminer. I have read through a few previous post but can't understand how to process this set of data.

I am trying to get it into a binary matrix where it will be like then apply the association rule.

Is there an easy way in Rapidminer 9.8 that can quickly transform this set of data into binary matrix or any other ways to preprocessed this kind of data for market basket analysis? If yes, how should I do it?

End result should be something like:
     yogurt   cheese roll....
C1    1               1
C2    1               0
C3    0               0
....
where x axis will be all the unique products in the basket
y axis is the customer ID

Thanks!

MartinLiebig

Hi @newjop228 ,

for awhile you do not need to change the format anymore, since we changed the FP-Grwoth operator to also work with list data like yours. The format you have is called 'item list in column'. It is the first parameter of Fp-Growth.

Cheers,

Martin

newjop228

Hey @mschmitz , thanks for the reply!

Cool that is great! I have tested it out but my results show nothing.

Operators:
Retrieve -> FP-Growth -> Create association rules -> output

Results:
FrequentItemSet(FP-growth) - no itemset found
AssociationRules(Create association rules) - No rules found

Any idea?

MartinLiebig

Hi,

did you try to lower the min_support setting of FP-Growth? This basically controls how strong the rule needs to be to find it. The lower the more results you get.

Cheers,

Martin

newjop228

Thanks @mschmitz. May I ask one last thing - for association rule, do I have to arrange the products to have an effective analysis? That is, column 1 will be "yogurt" and all customerID, column 2 will be "chicken". Or the association rule do not require me to arrange these products?

MartinLiebig

Hi,

order should not make a difference.

Best,

Martin