CART (Classification and Regression Trees)

The acronym CART is used in the data-mining industry both as (a) a generic for all methods of growing classification (unsupervised learning) and regression (supervised learning) trees and (b) a specific set of algorithms for growing binary trees introduced in the book by Breiman, Friedman, Olshen, and Stone (1984). The latter rely on sophisticated methods of cross-validation to adjust for problems introduced by the limitation of binary splits, which contrast with multisplit tree-growing algorithms such as CHAID (see description of CHAID).