[R] FSelector and RWeka problem

[Yanwei Song]

Hi,

Here is a follow up, I resolve the problem in another way. 
"
> e <- evaluate_Weka_classifier(p)
> print(subset)
> print(e$details[1])
> return(e$details[1])"

change this lines to 
"
e <- sum(iris$Species == predict(p))/nrow(iris)
print(subset)
print(e)
return(e)
================
It works well, but I still don't understand why the previous one didn't work. Maybe there is a compatibility problem...




Yanwei





On Sep 7, 2011, at 8:47 PM, Yanwei Song wrote:

> Hi all,
> 
> Last post didn't give the plain format:
> 
> I was trying to combine RWeka and FSelector, and do the forward feature selection with J48/C5.4 decision tree:
> 
> Here is the code:
> #==================
> library(RWeka)
> library(FSelector)
> library(rpart)
> 
> 
> data(iris)
> evaluator <- function(subset) {
> p <- J48(as.simple.formula(subset, "Species"), data=iris)
> e <- evaluate_Weka_classifier(p)
> print(subset)
> print(e$details[1])
> return(e$details[1])
> }
> 
> 
> subset <- forward.search(names(iris)[-5], evaluator)
> =========================
> I got this error, when I ran it:
> 
> Error in paste(attributes, sep = "", collapse = " + ") :
> cannot coerce type 'closure' to vector of type 'character'
> 
> I don't know how I could fix this problem.
> Thanks.
> 
> 
> Yanwei
> 
> 
> 
>