[Clus-general] Pruning of Random Forest

[Madeleine S. <mad...@in...>]

Hi,

I tried to run Clus using ensembles instead of a single tree. Using
random forests as ensemble method, I noticed that this setting does not
seem to work in combination with pruning. The output only contains two
models: the default model and the original model, but no pruning model.
My setting file looks as follows:

[General]
Verbose = 0

[Attributes]
Descriptive = 1-242
Target = 243-282
Clustering = 243-282

[Tree]
Heuristic = VarianceReduction
PruningMethod = C4.5

[Ensemble]
Iterations = 10
EnsembleMethod = RForest

[Data]
File = train.arff
TestSet = test.arff

Did I miss anything?

With regards to the single tree case, are there pruning methods other
than C4.5 that work in combination with classification?

Thank you and best regards,
Madeleine



-- 
Dipl.-Wirt.-Inf. Madeleine Seeland

Technische Universität München
Institut für Informatik Lehrstuhl I12 - Bioinformatik
Boltzmannstr. 3
85748 Garching b. München, Germany

Room: MI 01.09.040 Phone: +49 (89) 289-19443 
Email: mad...@in...
Web: http://wwwkramer.in.tum.de/people/seeland