mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nicholas Kolegraff <>
Subject Precision Recall -- Without RecommenderBuilder
Date Mon, 06 Feb 2012 20:33:58 GMT
Hi All,
I'm curious if there is a way to build evaluations of precision and recall
with only datasets.  It seems from the API that the
RecommenderIRStatsEvaluator requires you build a Recommender in mahout and
pass that to the evaluate method.
I am curious if something like this exists?

/* This code is only used to try and communicate what I mean */
DataModel trained = new FileDataModel( new File("path-to-file")) /* Already
trained dataset, containing predictions */
DataModel test = new FileDataModel( new File("path-to-file"))  /* Existing
testing data */
RecommenderIRStatsEvaluator eval = new RecommenderIRStatsEvaluator();
double score = eval.evaluate(trained, test)

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