Tree Building: Specifics

• Quinlan’s C4.5 (successor of ID3)

• Variables used in tree building

  • m = minimum number of cases needed in at least two outcomes of a tree node (test)
  • c = confidence level : used to find the upper limit on the probability of error at a leaf; used in pruning the tree
  • t = # of trees to be grown by partitioning the given training set; select best one

• Clinical judgment to help select the final tree (FT Tree)

Previous slide

Next slide

Back to first slide

View graphic version

17