You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@mahout.apache.org by ziad kamel <zi...@gmail.com> on 2012/04/03 00:57:02 UTC

How good recommendations and precision works

Hi ,  When not using a threshold, the default is to use average rating
plus one standard deviation. Assume that good recommendations was 30
items. When we use precision at 5 , we will select only top 5 items
from the 30 to compare with our recommendations ?

Re: How good recommendations and precision works

Posted by ziad kamel <zi...@gmail.com>.

Why we need a precision at and also threshold ? We can use only 1 of
them at a time ?

IRStatistics stats = evaluator.evaluate(recommenderBuilder,
                                            null, model, null, 5,

GenericRecommenderIRStatsEvaluator.CHOOSE_THRESHOLD,
                                            1.0);


On Mon, Apr 2, 2012 at 5:57 PM, ziad kamel <zi...@gmail.com> wrote:
> Hi ,  When not using a threshold, the default is to use average rating
> plus one standard deviation. Assume that good recommendations was 30
> items. When we use precision at 5 , we will select only top 5 items
> from the 30 to compare with our recommendations ?