ALSWR

[Han JU <ju...@gmail.com>]

Hi all,

I'm just getting my hands on Mahout to do some recommendation. I tried
SVDRecommender with ALSWRFactorizer on GroupLens data with boolean pref
datamodel, but the evaluation result seems to be always the same as the
lambda param.

For example if I use lambda=0.65, the average diff error and rmse will be
0.65. I tried several value from 0.02 to 0.8 and other datasets (a sample
from 1m songs) as well but this always happens. Some suggestions? The code
is very similar to that in the MiA book.

By the way, what I want is to use ALSWR/SVD in Mahout to process large
boolean pref datasets, like 1m songs, on Hadoop Clusters. Is this a good
idea?

Thanks!
-- 
*JU Han*

UTC   -  Université de Technologie de Compiègne
*     **GI06 - Fouille de Données et Décisionnel*

+33 0619608888