-3

I am building up a recommender system to recommend products that customers might be interested and I have a sales dataset, which has the customer ID and Product ID, and if the customer bought a product it would be represented by 1 or did not like the product would be represented by 0. Which is the best way to implement such use case using Mahout:

  1. If I am using User based recommender, what is the best user similarity and User Neighborhood
  2. What is the best way to evaluate this recommender system ?

Thank you !

user2255207
  • 41
  • 6
  • This question is off topic as it is primarily opinion based. – bhspencer Oct 03 '15 at 17:15
  • I won't agree that its opinion based as the performance and effectiveness of different similarity metrics can be evaluated for the user's specific data set. – Kris Oct 07 '15 at 07:39

1 Answers1

0

Have a look here for starters: Mahout Recommender Documentation. At the end of the page you will see how to evaluate your setup, you can of course pass there different recommender instances build with different similarity and neighbourhood you want to try out.

Kris
  • 5,714
  • 2
  • 27
  • 47