CS-ID3 (Tan & Schlimmer, 1989, 1990; Tan, 1993) is a TDIDT algorithm that selects the attribute that maximizes the following heuristic function:
We implemented CS-ID3 by modifying C4.5 so that it selects the attribute that maximizes (4).
CS-ID3 uses a lazy evaluation strategy. It only constructs the part of the decision tree that classifies the current case. We did not implement this aspect of CS-ID3, since it was not relevant for the experiments reported here.