It could still be that you're sampling too little test data for example. Or that the data is too sparse in some parts. This means there were 0 valid test cases.
On Thu, Oct 4, 2012 at 5:33 PM, yamo93 <[email protected]> wrote: > Strange, i have ~400.000 lines in my input file (~2500 users) ... > > > On 10/04/2012 05:52 PM, Sean Owen wrote: >> >> This is something fixed in trunk, but you can ignore it. It means your >> eval >> is on too little data and the result is not meaningful. >> On Oct 4, 2012 4:50 PM, "yamo93" <[email protected]> wrote: >> >>> Hi, >>> >>> I try to evaluate a boolean data model with RecommenderIRStatsEvaluator >>> (mahout 0.7). >>> >>> I always get this exception : >>> Exception in thread "main" java.lang.**IllegalArgumentException: Illegal >>> nDCG: NaN >>> at com.google.common.base.**Preconditions.checkArgument(** >>> Preconditions.java:92) >>> at org.apache.mahout.cf.taste.**impl.eval.IRStatisticsImpl.<** >>> init>(IRStatisticsImpl.java:**38) >>> at org.apache.mahout.cf.taste.**impl.eval.** >>> GenericRecommenderIRStatsEvalu**ator.evaluate(** >>> GenericRecommenderIRStatsEvalu**ator.java:212) >>> at net.bluekiwi.bluekarma.**recommender.Evaluate.main(** >>> Evaluate.java:36) >>> >>> I saw that an issue >>> (https://issues.apache.org/**jira/browse/MAHOUT-1015<https://issues.apache.org/jira/browse/MAHOUT-1015>) >>> >>> was fixed in this area, is it a collateral issue ? >>> >>> Rgds, >>> Yann. >>> >
