You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by Florian Geier <fl...@gmx.de> on 2010/02/19 22:18:20 UTC
recommendations over a subset of items
Hi everyone,
I'm just getting started with mahout and I was reading an old api
documentation that was referring to version 0.1. I noticed that in this
version it was possible to construct a Preference object with an instance of
the Item class. This class had a boolean property "recommendable". My
understanding was that this could be used to make use of the whole item
space for generating recommendations, even if only a subset of items was
actually recommendable (e.g. because the the other items are not on stock at
that time).
In version 0.2 this functionality seems to be no longer available - the
Preference constructor now only takes the IDs of users and items.
In my application I would have to make recommendations for a small subset of
items, but I would like to use all the items in the datamodel as input to my
collaborative filtering algorithm. Is there still a way to do that? If so,
it would be great if someone could give me a hint where I can find an
example, wiki post or tutorial on how to do it.
Regards,
Florian
--
View this message in context: http://old.nabble.com/recommendations-over-a-subset-of-items-tp27660374p27660374.html
Sent from the Mahout User List mailing list archive at Nabble.com.
Re: recommendations over a subset of items
Posted by Florian Geier <fl...@gmx.de>.
This seems to work perfectly.
Thanks a lot for your help!
srowen wrote:
>
> Yes, the equivalent now would be to use a Rescorer object, which would
> return false from isFiltered() for an item that should not be
> recommended. How's that work for you?
>
> On Fri, Feb 19, 2010 at 9:18 PM, Florian Geier <fl...@gmx.de>
> wrote:
>>
>> Hi everyone,
>>
>> I'm just getting started with mahout and I was reading an old api
>> documentation that was referring to version 0.1. I noticed that in this
>> version it was possible to construct a Preference object with an instance
>> of
>> the Item class. This class had a boolean property "recommendable". My
>> understanding was that this could be used to make use of the whole item
>> space for generating recommendations, even if only a subset of items was
>> actually recommendable (e.g. because the the other items are not on stock
>> at
>> that time).
>> In version 0.2 this functionality seems to be no longer available - the
>> Preference constructor now only takes the IDs of users and items.
>>
>> In my application I would have to make recommendations for a small subset
>> of
>> items, but I would like to use all the items in the datamodel as input to
>> my
>> collaborative filtering algorithm. Is there still a way to do that? If
>> so,
>> it would be great if someone could give me a hint where I can find an
>> example, wiki post or tutorial on how to do it.
>>
>> Regards,
>> Florian
>> --
>> View this message in context:
>> http://old.nabble.com/recommendations-over-a-subset-of-items-tp27660374p27660374.html
>> Sent from the Mahout User List mailing list archive at Nabble.com.
>>
>>
>
>
--
View this message in context: http://old.nabble.com/recommendations-over-a-subset-of-items-tp27660374p27672244.html
Sent from the Mahout User List mailing list archive at Nabble.com.
Re: recommendations over a subset of items
Posted by Sean Owen <sr...@gmail.com>.
Yes, the equivalent now would be to use a Rescorer object, which would
return false from isFiltered() for an item that should not be
recommended. How's that work for you?
On Fri, Feb 19, 2010 at 9:18 PM, Florian Geier <fl...@gmx.de> wrote:
>
> Hi everyone,
>
> I'm just getting started with mahout and I was reading an old api
> documentation that was referring to version 0.1. I noticed that in this
> version it was possible to construct a Preference object with an instance of
> the Item class. This class had a boolean property "recommendable". My
> understanding was that this could be used to make use of the whole item
> space for generating recommendations, even if only a subset of items was
> actually recommendable (e.g. because the the other items are not on stock at
> that time).
> In version 0.2 this functionality seems to be no longer available - the
> Preference constructor now only takes the IDs of users and items.
>
> In my application I would have to make recommendations for a small subset of
> items, but I would like to use all the items in the datamodel as input to my
> collaborative filtering algorithm. Is there still a way to do that? If so,
> it would be great if someone could give me a hint where I can find an
> example, wiki post or tutorial on how to do it.
>
> Regards,
> Florian
> --
> View this message in context: http://old.nabble.com/recommendations-over-a-subset-of-items-tp27660374p27660374.html
> Sent from the Mahout User List mailing list archive at Nabble.com.
>
>