Index of TwoStageCollective_20050728

 
Setup
=====

- split (train/test): 
  ~ weka27a2:  5/95%
  ~ weka27b:  10/90%
  ~ weka27c:  20/80%
- repetitions: 10
- TwoStageCollective with only one iteration, i.e., no flipping takes place
  ~ first classifier trains on training set, labels the test set
  ~ second classifier train on the pre-labeled test set and tries to predict
    the original training set
- different classifiers tested


Evaluation
==========

SMO and NaiveBayes perform better, i.e., fewer losses, with the smaller
training set size (split 5/95%). Most of the other setups have only a few changes (1-2 wins/losses).
 
graphs(1.1M)
setup(180K)
stats(224K)