2 Replies Latest reply: Sep 9, 2012 2:56 AM by 769738 RSS

    Association technique to make product recommendation

    769738
      Hi Everyone,

      I would like to do data mining analysis using association technique. The data is sales transaction of automotive company that sale many different kind of vehicle. I know Association technique usually used for market basket analysis to understand purchase behavior of customers or to discover what kind of product that often to be purchased together.

      My Question is : Can I use association technique to analyze purchase behaviour based on repeat order data but in different transaction not at the same time. For example :

      Customer A bought Vehicle Type 1 at *1 Jan 2012*
      Customer A bought Vehicle Type 2 at *6 Jun 2012*
      Customer B bought Vehicle Type 3 at *7 Jun 2012*
      ...
      ...

      The objective is to make product recommendation to customers based on the patterns. Is it OK to use association technique for this situation, any Suggestion ?

      Thanks
        • 1. Re: Association technique to make product recommendation
          604934
          Yes definitely you can use. Simplest method maybe to use pivoting and decision trees together. Let you have the data
          Customer A bought Vehicle Type 1 at 1 Jan 2012
          Customer A bought Vehicle Type 2 at 6 Jun 2012
          Customer A bought Vehicle Type 3 at 7 Jun 2012
          Customer B bought Vehicle Type 1 at 5 Jun 2012
          Customer B bought Vehicle Type 3 at 7 Jun 2012

          You can convert this transactional data into
          Lag1 Lag2 Target
          Type2 Type1 Type3 -- For customer A
          Type2 NULL Type3 -- For customer A
          Type1 NULL Type3 -- For customer B
          .
          .

          When you give target column as "Target" to ODM you will obtain several decision tree rules. You need to verify the power of those rules over a test set.

          Or you can use arule package with Oracle Enterprise R which you may find easier.
          • 2. Re: Association technique to make product recommendation
            769738
            OK thanks for your suggestion, I'll check on it