Several base classifiers (J48, Logistic, RandomForest, SMO, Adaboost/J48, IB1)
are tested against various YATSI setups with the same base classifier.
Four different weighting setups:
- no-weights
- weight of 0.1
- weight of 1.0
- weight of 10.0
Four different (random) train/test splits:
- 1/99%
- 5/95%
- 10/90%
- 20/80%
The 1/99% split is run with different numbers of neighbors:
- 10 - weka39a
- 20 - weka39a2
- 50 - weka39a3
Iterations: 20 |