Describes a tree-based classifier. It handles un-ordered domains and null/missing values. It uses entropy arguments to choose attributes on which to discriminate. Interesting points: 1. pointer to a db repository: ml-repositoryics.uci.edu (UC Irvine) 2. comparison with statistics-based systems (p. 15) "As a general rule ... statistical techniques tend to focus on tasks in which all the attributes have continuous or ordinal values" S.M. Weiss and C.A. Kulikowski "Computer Systems that Learn", Morgan Kaufmann, San Mateo CA, 1991 provide a comparison 3. MDL principle, to decide how to prune the tree Jorman Rissanen, Anal of Statistics, 11,2, 416-431 Quinlan and Rivest, "Inferring decision trees using the Minimum Description Length principle" Information and Computation 80,3, 227-248 4. comparison of tree-based classifiers with neural nets (NN) - they are both more robust - they are about equally accurate (with NN slightly ahead); but NN require much more computation (an order of magnitude more) (p. 102) 5. CART is a statistics-based program L. Breiman, J.H. Friedman, R.A. Olshen and C.J. Stone "Classification and Regression Trees" Belmont, CA: Wadsworth (1984) 6. citation: Hunt 75 "Artificial Intelligence" NY, Academic Press (pioneered the tree-based classification methods)
CITATION STYLE
Quinlan, J. R. (1993). {C4}.5 - Programs for Machine Learning. San Mateo: Morgan Kaufmann.
Mendeley helps you to discover research relevant for your work.