About recommenders and user data

32 views
Skip to first unread message

Armando Pérez Marqués

unread,
Nov 25, 2011, 2:59:41 PM11/25/11
to Scikit-Crab
Hi! I'm new to the concept of recommender engines, I'll try to express
my best.

I'm buiding a site which sells certaint type of product, and one of
the requirements is to provide good recommendations for options
similar to the product the visitor requested, in case that product is
not available at the moment.

I want to provide recommendations, initially there will be no user
data/ratings on any product. I have a classification system to
classify the products in several categories, and every category has a
weight, to provide info on what category is more important to evaluate
similarities. Also I want to include other parameters in the
evaluation, like price and deliver time (not part of the category
system).

The cuestion is, can I (right now) use Crab for this? If
(unfortunately) no, there's something prepackaged I can use for this
or I have to roll my own? I can happily provide more details.

Greetings,

Armando

Marcel Caraciolo

unread,
Nov 29, 2011, 11:24:13 PM11/29/11
to Scikit-Crab
Hi Armando,
How are you ? Sorry for delay, my e-mail inbox wasn't receiving e-mails about crab.

As I understood your problem you want to build a recommender engine with some constraint parameters (for instance
 the features of the similar product) . What  happens is that those type of recommenders are known as Knowledge-Based recommenders. The
recommendations are calculated independently of individual ratings (no ratings required) in the form of similarities between
the customer requirements and items.  You should take a deep look at the article attached about constraint recommenders.

Crab still does not implement Knowledge Recommenders yet. But you could implement your own using the baseline classes we provide for construct your own recommender systems.  I don't know any knowledge recommender project open-source available. But we can help you to make this happen with crab.

Let me know your requirements and what you need to start.

Marcel Caraciolo


I want to provide recommendations, initially there will be no user
data/ratings on any product. I have a classification system to
classify the products in several categories, and every category has a
weight, to provide info on what category is more important to evaluate
similarities. Also I want to include other parameters in the
evaluation, like price and deliver time (not part of the category
system).

The cuestion is, can I (right now) use Crab for this? If
(unfortunately) no, there's something prepackaged I can use for this
or I have to roll my own? I can happily provide more details.
--
Marcel Pinheiro Caraciolo
M.S.C. Candidate at CIN/UFPE
a3-felfernig.pdf
Reply all
Reply to author
Forward
0 new messages