You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by Seby Paul <se...@yahoo.com> on 2011/02/04 23:12:18 UTC
mahout recommendation help
Hi,
There is no user preference in our data file. i can create recommendations with mhaout using LogLikelihoodSimilarity and TanimotoCoefficientSimilarity.
Is it possible to use any other similarity metrics without the preference value ?
thank you
seby paul
Re: mahout recommendation help
Posted by Sean Owen <sr...@gmail.com>.
There's no built-in function but you could iterate through the data model
and compute that easily.
On Sun, Feb 6, 2011 at 5:46 PM, Steven Bourke <st...@ucd.ie> wrote:
> Yeah if something is to good to be true... ! I didn't notice anything
> particularly different about the item based approach over the user based.
>
> Are there functions built into mahout so I can check how dense / sparse
> the user,item matrix is?
>
>
Re: mahout recommendation help
Posted by Steven Bourke <st...@ucd.ie>.
Yeah if something is to good to be true... ! I didn't notice anything particularly different about the item based approach over the user based.
Are there functions built into mahout so I can check how dense / sparse the user,item matrix is?
Steven Bourke
Clarity Centre for Sensor Web Technologies,
Science North,
School of Computer Science & Informatics,
University College Dublin.
On 6 Feb 2011, at 17:35, Sean Owen wrote:
> I don't know of anything of particular significance. Good results are good,
> but, too good suggests a target leak or something.
>
> On Sun, Feb 6, 2011 at 5:15 PM, Steven Bourke <st...@ucd.ie> wrote:
>
>> Has anyone used the boolean pref item recommender from 0.5 snapshot? I was
>> wondering if there is any other work being carried out on it ? I couldn't
>> find anything in the JIRA system.
>>
>> I ask because I ran it on some data as a comparison and got surprisingly
>> good results when running evaluations against it.
>>
>>
Re: mahout recommendation help
Posted by Sean Owen <sr...@gmail.com>.
I don't know of anything of particular significance. Good results are good,
but, too good suggests a target leak or something.
On Sun, Feb 6, 2011 at 5:15 PM, Steven Bourke <st...@ucd.ie> wrote:
> Has anyone used the boolean pref item recommender from 0.5 snapshot? I was
> wondering if there is any other work being carried out on it ? I couldn't
> find anything in the JIRA system.
>
> I ask because I ran it on some data as a comparison and got surprisingly
> good results when running evaluations against it.
>
>
Re: mahout recommendation help
Posted by Steven Bourke <st...@ucd.ie>.
Has anyone used the boolean pref item recommender from 0.5 snapshot? I was wondering if there is any other work being carried out on it ? I couldn't find anything in the JIRA system.
I ask because I ran it on some data as a comparison and got surprisingly good results when running evaluations against it.
Steven Bourke
Clarity Centre for Sensor Web Technologies,
Science North,
School of Computer Science & Informatics,
University College Dublin.
On 5 Feb 2011, at 23:11, Lance Norskog wrote:
> Hi-
>
> This is a boolean case, so you can use the Boolean preference and
> similarity classes.
>
> On Fri, Feb 4, 2011 at 2:12 PM, Seby Paul <se...@yahoo.com> wrote:
>> Hi,
>> There is no user preference in our data file. i can create recommendations with mhaout using LogLikelihoodSimilarity and TanimotoCoefficientSimilarity.
>>
>> Is it possible to use any other similarity metrics without the preference value ?
>>
>>
>> thank you
>> seby paul
>>
>>
>>
>
>
>
> --
> Lance Norskog
> goksron@gmail.com
Re: mahout recommendation help
Posted by Lance Norskog <go...@gmail.com>.
Hi-
This is a boolean case, so you can use the Boolean preference and
similarity classes.
On Fri, Feb 4, 2011 at 2:12 PM, Seby Paul <se...@yahoo.com> wrote:
> Hi,
> There is no user preference in our data file. i can create recommendations with mhaout using LogLikelihoodSimilarity and TanimotoCoefficientSimilarity.
>
> Is it possible to use any other similarity metrics without the preference value ?
>
>
> thank you
> seby paul
>
>
>
--
Lance Norskog
goksron@gmail.com