Oct 22, 2003 ------------- - More on overfitting - why it happens - example from PlayTennis - Troublesome datasets for decision trees - parity function - majority function - m-of-n functions - Problems with decision tree learning - fragmentation - use gain ratio instead of gain - duplication of subtrees - arises because of greedy algorithm - Twists to the tale - what to do with continuous-valued attributes? - binning strategies - what to do with missing attributes? - use probabilistic interpretations - read the book