Recommender systems evaluation

[Hovnatan Karapetyan <ho...@gmail.com>]

Hi,

I am a new Mahout user, trying to build my first recommendation
engine. I'm using mahout-0.9. My data is boolean -- user, item
association without any rating. I am using
GenericBooleanPrefUserBasedRecommender with
TanimotoCoefficientSimilarity. When I evaluate the recommendation
engine with GenericRecommenderIRStatsEvaluator, it starts outputting
the evaluation results for each individual user. After the evaluation
finishes I run getPrecision() on the return (IRStatistics) and it
shows the precision value of the last individual user recommendation
precision. Shouldn't it instead return some sort of average of the all
individual precision values not the last user value? Also it seems
that for all cases precision and recall values are equal, why is that?
Is there anything I should do that the evaluation process runs in
multithreaded way or is this automatically the case?

Thanks!
Hovnatan