On Wed, 3 Jun 2009, Huihua Lu wrote:
> Hi everyone,
>
> I have trouble to use RWeka, I tried: (w=weather dataset, all
> preditors are nominal)
>
>> m<-J48(play~., data=w)
>> e<-evaluate_Weka_classifier(m,cost = matrix(c(0,2,1,0),
> + ncol = 2),numFolds = 10, complexity = TRUE,seed = 123,
> + class = TRUE)
> it gives me exactly what I want, but when I tried the same classifier
> on the other published data: (iris dataset has all numeric preditors)
>
>> m<-J48(Species~., data=iris)
>> e <- evaluate_Weka_classifier(m,
> + cost = matrix(c(0,2,1,0), ncol = 2),
> + numFolds = 10, complexity=TRUE, class=TRUE)
> it threw me the error as below:
>
> Error in .jnew("weka/classifiers/Evaluation", instances,
costMatrix) :
> Failed to create object of class `weka/classifiers/Evaluation'
> In addition: Warning message:
> In .jnew("weka/classifiers/Evaluation", instances, costMatrix) :
> NewObject
>
("weka/classifiers/Evaluation","(Lweka/core/Instances;Lweka/classifiers
> /CostMatrix;)V",...) failed
>
> I know there are something wrong in the function of
> evaluate_Weka_classifier, but I am stuck. Any one can help me out?
> Is it because the attributes of predictors? then, how can I set up
> the parameters in evaluat_Weka_classifier function? Thanks very much!
For me, it throws one more error, that you didn't include:
Exception in thread "main" java.lang.Exception: Cost matrix not
compatible
with data!
at weka.classifiers.Evaluation.<init>(Unknown Source)
Your response has three levels, hence the cost matrix needs to be 3x3.
e <- evaluate_Weka_classifier(m,
cost = matrix(c(0, 1, 2, 2, 0, 1, 1, 1, 0), ncol = 3),
numFolds = 10, complexity=TRUE, class=TRUE)
works correctly but throws a warning. This warning is not critical,
though, and should be fixed in the next RWeka version.
Best,
Z
> lu
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
>