Iterative Dichotomizer 3
- tries to create a shallow tree
 - greedy
 - recursive depth first
 
Algorithm
- choose best feature using Information Gain Ratio or some other metric, dont pick this feature again.
 - partition dataset using selected feature (all possible branches)
 - repeat until break condition
 
Break condition
- homogenous or pure set → mark as classfication
 - no information gain if split
 - run out of features to split on