Product Pairing

ritsard_marzan

Hi. Can someone please help me about using RapidMiner for the following:

I have around 2000+ products which customers select and buy.

Sales transactions are recorded in a sales table which is linked to a sales details table where items per sales transactions are recorded and where products actually bought are saved (together with the quantity, price, discount etc.)

 

My problem is this: I want to use RapidMiner in order for me to find out the top pair of products bought by the customers for a specific period.
Example:

If Customer1 buys Product1 and Product2, then the first Product Pairing is:

     Pair1(Product1,Product2,Count=1);

 

If Customer 2 buys Product3 and Product4, then the product pairing will be Pair(Product3,Product4). It will look into the Product Pairing record if it's already there, if not yet, then add it there. so we now have 2 pairs:

     Pair1(Product1,Product2,Count=1);
     Pair2(Product3,Product4,Count=1);

 

If Customer 3 buys Product1 and Product2, since we already have that pairing, it will only increase the count, thus:

     Pair1(Product1,Product2,Count=2);
     Pair2(Product3,Product4,Count=1);

 

If Customer 4 buys Product1, Product2, and Product3, then we have 3 pairs here: (Product1,Product2); (Product1,Product3); and (Product2,Product3), thus our pairing records will be:

     Pair1(Product1,Product2,Count=3);

     Pair2(Product3,Product4,Count=1);

     Pair3(Product1,Product3,Count=1);
     Pair4(Product2,Product3,Count=1);
    

 

With the following, we can say that the product pair mostly bought by customers is Pair1(Product1,Product2).

Can this be possible in RapidMiner? I have a storedProdcedure in MySQL that can do this but since there are thousands of sales transactions having thousands of sales details, it's taking a lot of time and resources in server resources.

I hope somebody could link me to possible PDF, tutorial, or weblink that will guide me on this.

Thanks in advance.

Thomas_Ott

Completely and easily doable in RapidMiner.

 

Do a search for FP-growth and/or Association Rules in the forum. There's a lot of sample processes and applications to this type of Market Basket analysis. 

ritsard_marzan

thanks for the swift reply. will surely try your suggestions