Scale content recommendation

This question is probably very repeated on blog sites and Q & A, but I have not yet found a specific answer.

I am trying to create a recommendation system for customers using only their purchase history.

  • Say my application has n products.
  • Calculate the similarities of elements for all n products based on their attributes (for example, country, type, price)
  • When the user needs a recommendation, loop the previously purchased products p for user u and select similar products (the similarity is performed in the previous step).

If we are right, we call it a recommendation based on content, and not joint filtering, since it is not related to the coincidence of elements or user settings with the element.

My problem is multiple:

  • Is there any existing scalable ML platform that solves competition-based recommendations (I'm fine to use different technologies / language).
  • Is there a way to tune Mahout to get this result?
  • Is classification a way to handle content-based recommendations?
  • Is this something suitable for this task?

Note. I looked at Mahout (since I am familiar with Java and Mahout seems to use Hadoop for distributed processing) for the scale and benefits of having well-tested ML algorithms.

Your help is appreciated. Any examples would be really wonderful. Thank.

+4
source share
1 answer

, . , , K , , matix .

, Mahout: Mahout ( )?

: , .

, , : -

0

Source: https://habr.com/ru/post/1584352/


All Articles