Friday, March 03, 2006

Recommendation and Rating Engines

I am researching recommender and rating systems for two clients at the same time for completely different reasons. One client wants to be able to be at a single or list of products or categories and recommend similiar or logical cross selling opportunities. The other client wants to be able to allow a user to define a setting that is ok and then give the user alternate settings that are algorithmically similiar but possibly superior to the original settings.

Here are some things I have run into while taking a quick look on the web.
(Common terminology: Recommendation Engine, Recommender Engine, Collaborative Filtering)

If you are trying to get targeted recommendations and cross selling products these two products are interesting to look at:
SUGGEST Karypis lab at the Univ. of MN (like Amazon's, "You might also like" engine)

If you are trying to build a user defined rating of products or services take a look at:
CoFE IIS Research Group of Oregon State(like Yahoo's news article ranking system)

Often companies will combine these two technologies to allow users to refine the products and services recommended to them.

For example: A baker views flour and sugar and is recommended yeast, flavorings and preservatives. Let's say the baker is an all natural baker and can rate the preservatives as something he would never use.

To see a site that uses both go to
The home page has recommendations based on user ratings. If you click on a specific artist or album you will see a list of Similar Music. This site uses a content based recommendation engine from Loomia.

1 comment:

Replacement Parts said...

Wow It really seems like mister here did his homework! Great effort in constructing this information, I really got a new sense of what's going on. Well, at least we can call it an update to what I already knew. But it's good to have yourself really looking into this stuff and putting it out there. I probably would have not investigated this material in wow g-d knows how long. I can really say I picked up some knowledge from this piece and you can most definitely count me referring some readers to your page, thanks! Really enjoyed and it was my pleasure to read on!