Hi Stevo,

So the method predictFromFactorization, which was in PredictorJob, seems
not to have been migrated over.  RecommenderJob gives me top N
recommendations (plus predicted preferences).  predictFromFactorization is
handy because I can pass in userid, itemid pairs and get a preference
prediction, such as the requirement from the Netflix challenge.  It seems
this feature was only in 0.5.

Regards

Stuart




On 13 September 2013 05:53, Stevo Slavić <[email protected]> wrote:

> Hello Stuart,
>
> Have a look at description of
> https://issues.apache.org/jira/browse/MAHOUT-872
> In a comment there Sebastian also references wiki entry:
>
> https://cwiki.apache.org/confluence/display/MAHOUT/Collaborative+Filtering+with+AlS-WR
>
> PredictionJob/predictFromFactorization was replaced with
> org.apache.mahout.cf.taste.hadoop.als.RecommenderJob = recommendfactorized
> : Compute recommendations using the factorization of a rating matrix
>
> See the RecommenderJob class javadoc for parameters:
>
> https://builds.apache.org/job/Mahout-Quality/javadoc/org/apache/mahout/cf/taste/hadoop/als/RecommenderJob.html
>
> Kind regards,
> Stevo Slavic.
>
>
> On Thu, Sep 12, 2013 at 11:49 AM, Stuart Horsman
> <[email protected]>wrote:
>
> > So I'm using predictFromFactorization in Mahout 0.5 but this code was
> > removed from 0.6. Is there any special reason for this?
> >
> > Thanks
> >
> > Stuart
> >
> >
> > On 12 September 2013 08:28, Stuart Horsman <[email protected]>
> > wrote:
> >
> > > Hi All,
> > >
> > > I'm new to mahout so thanks up front for the help.  I'm running ALS-WR
> on
> > > a sparse matrix of movies and items and getting some recommendations
> out.
> > >  However I want to predict user preference.  I noticed there was a
> > predictFromFactorization
> > > in previous versions but seems to have been removed.
> > >
> > > What's the best approach I should take using 0.8?
> > >
> > > Thanks
> > > Stuart
> > >
> >
>

Reply via email to