Yes, it looks like something strange is going on. I'll take a closer
look when I get a chance. Thanks very much for giving us access to the
data! (As far as I can tell, the only aspect that makes the data
special are some attributes that have extremely large values. Maybe the
problem is related to that.)
On Aug 17, 2005, at 10:57 PM, juan jose rodriguez diez wrote:
With the following data set:
I get a "StackOverflowError" with J48.
- There is no problem with the data set in Quinlan's C4.5
- If I normalize the data set, J48 has no problems.
I think it is a case of infinite recursion.
Juan José Rodríguez, http://pisuerga.inf.ubu.es/juanjo
Lenguajes y Sistemas Informáticos, Universidad de Burgos
Wekalist mailing list