You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by EDUARDO ANTONIO BUITRAGO ZAPATA <ed...@gmail.com> on 2012/05/12 03:15:25 UTC
Recommender with item features
Hi All,
As far I've seen, all mahout recommenders uses this setting for input file:
User1, item1, rating1
User1, item2, rating2
...
User2, item1, rating3
What I need to do is a recommender for new digital cameras. I want to know
which user is more interested in a camera when it arrives, then I can make
recommendation. To do that, I want to take into account the camera
features (optical zoom, LCD size, etc). ¿Is there a way to implement this
in mahout? Maybe the input file is something like this:
User1, item1, feat1, feat2, … , featn, rating1
User1, item2, feat1, feat2, … , featN, rating2
...
User1, itemN, feat1, feat2, … , featN, rating3
User1, newItem, feat1, feat2, … , featn, ?
Any help would be appreciated.
--
Eduardo** <ea...@uniandes.edu.co>
Re: Recommender with item features
Posted by EDUARDO ANTONIO BUITRAGO ZAPATA <ed...@gmail.com>.
Some features are numeric (like LCD size screen, resolution, optical zoom)
and others are categorical (like camera type, image sensor type). I’m
working from data that I’ve scrapped from epinions.com (academic purposes)
but I haven’t normalized yet. Now for the ItemSimilarity implementation,
maybe I would use rochio algorithm but I’m not sure because I need to
understand it better or understand others algorithms.
Is good to know that mahout can be useful for this problem, thanks for
your reply.
2012/5/12 Sean Owen <sr...@gmail.com>
> You can write your own ItemSimilarity metric based on the features and
> then use an item-based recommender. That piece you'd have to do
> yourself by making up some notion of similarity; if the features were
> all numeric and normalized you can look at repurposing something based
> on Euclidean distance or cosine similarity or something, but I don't
> know if those are the features you have.
>
> Sean
>
> On Sat, May 12, 2012 at 2:15 AM, EDUARDO ANTONIO BUITRAGO ZAPATA
> <ed...@gmail.com> wrote:
> > Hi All,
> >
> > As far I've seen, all mahout recommenders uses this setting for input
> file:
> >
> > User1, item1, rating1
> > User1, item2, rating2
> > ...
> > User2, item1, rating3
> >
> > What I need to do is a recommender for new digital cameras. I want to
> know
> > which user is more interested in a camera when it arrives, then I can
> make
> > recommendation. To do that, I want to take into account the camera
> > features (optical zoom, LCD size, etc). ¿Is there a way to implement
> this
> > in mahout? Maybe the input file is something like this:
> >
> > User1, item1, feat1, feat2, … , featn, rating1
> > User1, item2, feat1, feat2, … , featN, rating2
> > ...
> > User1, itemN, feat1, feat2, … , featN, rating3
> > User1, newItem, feat1, feat2, … , featn, ?
> >
> > Any help would be appreciated.
> >
> > --
> > Eduardo** <ea...@uniandes.edu.co>
>
--
EDUARDO BUITRAGO
Est. Msc. en Ingeniería - Sistemas y Computación - Universidad de los Andes
Ing. de Sistemas - Universidad Francisco de Paula Santander
Cisco Certified Network Associate - CCNA
Re: Recommender with item features
Posted by Sean Owen <sr...@gmail.com>.
You can write your own ItemSimilarity metric based on the features and
then use an item-based recommender. That piece you'd have to do
yourself by making up some notion of similarity; if the features were
all numeric and normalized you can look at repurposing something based
on Euclidean distance or cosine similarity or something, but I don't
know if those are the features you have.
Sean
On Sat, May 12, 2012 at 2:15 AM, EDUARDO ANTONIO BUITRAGO ZAPATA
<ed...@gmail.com> wrote:
> Hi All,
>
> As far I've seen, all mahout recommenders uses this setting for input file:
>
> User1, item1, rating1
> User1, item2, rating2
> ...
> User2, item1, rating3
>
> What I need to do is a recommender for new digital cameras. I want to know
> which user is more interested in a camera when it arrives, then I can make
> recommendation. To do that, I want to take into account the camera
> features (optical zoom, LCD size, etc). ¿Is there a way to implement this
> in mahout? Maybe the input file is something like this:
>
> User1, item1, feat1, feat2, … , featn, rating1
> User1, item2, feat1, feat2, … , featN, rating2
> ...
> User1, itemN, feat1, feat2, … , featN, rating3
> User1, newItem, feat1, feat2, … , featn, ?
>
> Any help would be appreciated.
>
> --
> Eduardo** <ea...@uniandes.edu.co>