mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Charly Lizarralde <charly.lizarra...@gmail.com>
Subject Re: Estimating preferences with Distributed Item Based Recommender
Date Fri, 05 Oct 2012 14:05:51 GMT
Of course.

I used the DataSetSplitter for spliting the datasets into training & probe.

I created a job very similar to the aggregateAndRecommender step of the
RecommenderJob. I copied and modified the
AggregateAndRecommendReducer.class to be EstimatePreferenceReducer.class
that loads in the setup the probe dataset.

If you want I can send you the code...but it's not very clean.

On Fri, Oct 5, 2012 at 10:42 AM, Matt Mitchell <goodieboy@gmail.com> wrote:

> Hi Charly. Would you mind showing, or at least describing how you did this?
>
> Thanks,
> Matt
>
> On Fri, Oct 5, 2012 at 9:21 AM, Charly Lizarralde
> <charly.lizarralde@gmail.com> wrote:
> > Finally I could manage to do that. It was not so difficult after all. I
> > used that for calculating RMSE.
> >
> > Thanks!
> > Charly
> >
> > On Wed, Oct 3, 2012 at 1:54 PM, Charly Lizarralde <
> > charly.lizarralde@gmail.com> wrote:
> >
> >> Thanks! I think I'll manage.
> >>
> >> On Wed, Oct 3, 2012 at 11:47 AM, Sean Owen <srowen@gmail.com> wrote:
> >>
> >>> Yes... you'd have to modify it a little though. The stage right before
> >>> recommendation in the pipeline computes all the estimates, on which
> >>> the results are ranked. You would just change this to output these
> >>> all, or, just the ones you want. There's nothing out of the box that
> >>> does this.
> >>>
> >>> On Wed, Oct 3, 2012 at 2:53 PM, Charly Lizarralde
> >>> <charly.lizarralde@gmail.com> wrote:
> >>> > Is it possible to use RecommenderJob (Map Reduce/Item Based ) to
> >>> estimate
> >>> > preferences por specific <user,item> dataset ?
> >>> >
> >>> > Thanks!
> >>> > Charly
> >>>
> >>
> >>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message