Question for Collaborative Filtering with ALS-WR

[Qi Zhou <qz...@virginia.edu>]

Hi,

I am a student at University of Virginia. I am using ALS-WR in Mahout and
encounter some questions. So I have the command "-parallelALS" running, and
I want to use the matrix to predict rating. I am using 0.7 version and can
only compute RMSE. Is that possible that I get the predicted rating from
the matrix?

Any of your help will be really appreciated!


Best Regards,
Qi Zhou

Re: Question for Collaborative Filtering with ALS-WR

[Sebastian Schelter <ss...@apache.org>]

Hi Qi,

its absolutely possible. Have a look at PredictRatingsMapper in
org.apache.mahout.cf.taste.hadoop.als.FactorizationEvaluator

This mapper loads the feature matrices U and M into memory and predicts
ratings by computing dot products. Afterwards it only saves the error
but you should be able to modify the code to save predicted ratings.

Best,
Sebastian

On 19.03.2013 12:17, Qi Zhou wrote:
> Hi,
> 
> I am a student at University of Virginia. I am using ALS-WR in Mahout and
> encounter some questions. So I have the command "-parallelALS" running, and
> I want to use the matrix to predict rating. I am using 0.7 version and can
> only compute RMSE. Is that possible that I get the predicted rating from
> the matrix?
> 
> Any of your help will be really appreciated!
> 
> 
> Best Regards,
> Qi Zhou
>