2

So, I'm brainstorming up a recommended-item algorithm, and I like Amazon's. It has "Customers who bought this also bought" and "related" recommended sections. I'm guessing the former is based on sales data, and the other is based on catalog data, such as category, attributes, etc.

Does anyone know how Amazon or other similar large retail companies pull these recommended products onto their product pages? Is it done on the fly or done at certain intervals? I can't decide which way it should be done at this time. I'm leaning toward the latter, because the algorithm can be more intricate, but I'm guessing this will need to be processed and updated at certain interval to minimize page load times.

Any comments from programmer who have worked on this type of project are welcome!

laketuna
  • 3,832
  • 14
  • 59
  • 104
  • 1
    There is a great answer in here that covers this - http://stackoverflow.com/questions/2323768/how-does-the-amazon-recommendation-feature-work - also there is a new xcommerce module for magento which may be of interest to you: http://www.magentocommerce.com/magento-connect/product-recommendations.html – Drew Hunter Jun 12 '12 at 19:25

2 Answers2

1

Amazon use something called item based collaborative filtering.

Google "amazon item based collaborative filtering" as I can't copy paste hte link due to Google's sneaky tracking!

Steve
  • 21,163
  • 21
  • 69
  • 92
0

My team constructed a recommender system using collaborative filtering and we had an offline pre-processing phase and another phase that was calculated on the fly. The preprocessing phase involved the most expensive calculation, so the final part could be done easily and fast. We used the SlopeOne algorithm, which is already structured like that.

Amazon also uses collaboartive filtering, described here:

http://www.cs.umd.edu/~samir/498/Amazon-Recommendations.pdf

Community
  • 1
  • 1
Renata Ghisloti
  • 547
  • 6
  • 13