Next: Reduced Error Pruning
Up: Decision Tree Learning
Previous: Overfitting in Decision Trees
- stop growing tree earlier
- post-prune the tree
- separate set of examples - training and validation set approach
- even if the training set is mislead by random errors the validation
set is unlikely to exhibit the same random fluctuations - 2/3
training, 1/3 validation
- statistical test
- measure for complexity of encoding
Patricia Riddle
Fri May 15 13:00:36 NZST 1998