4
$\begingroup$

I am trying to build a recommender system for a supermarket and do not really have any ratings for any product apart from the number of purchases of each item. A sample of the data is like:

KL_ID     prod1     prod2     prod3     prod4     prod5     prod6     prod7     prod8
 1839        18         0         0         0        23         0         0         0
 1943         0         0         0         0         0         0         0         0
  632         0         0         0         0         2         0         0        30
 4852         0         0         0         0        52         0         0         6

Based on this data I am trying to predict the new items each KL_ID will buy. Since I don't have any ratings, I can't do it the usual way. An alternative can be to normalize this and assume it is a rating and then use recommenderlab.

$\endgroup$
  • 1
    $\begingroup$ recommend products that people usually buy together? $\endgroup$ – cory Mar 14 '16 at 14:39
  • 1
    $\begingroup$ you probably should google for "implicit feedback" - see here aaai.org/Papers/Workshops/1998/WS-98-08/WS98-08-021.pdf $\endgroup$ – dratewka Mar 14 '16 at 15:00
  • 1
    $\begingroup$ @cory : recommend products that each KL_ID might be interested in based on previous history and comparison with other KL_IDs who purchased similar products. I hope its clear? $\endgroup$ – Prashanth Mar 14 '16 at 15:11
3
$\begingroup$

Alternative to your approach, you can use Sequential pattern mining algorithms if you have timestamps of purchases. Or using some algorithm for association rules if you don't. SPMF(www.philippe-fournier-viger.com/spmf/) Library has both of these types of algorithms.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy