Hierarchical priors for bayesian CART shrinkage

被引:10
作者
Chipman, H
George, EI
McCulloch, RE
机构
[1] Univ Waterloo, Dept Stat & Actuarial Sci, Waterloo, ON N2L 3G1, Canada
[2] Univ Texas, Dept MSIS, Austin, TX 78712 USA
[3] Univ Chicago, Grad Sch Business, Chicago, IL 60637 USA
基金
美国国家科学基金会;
关键词
binary trees; tree shrinkage; Markov chain Monte Carlo; model selection; stochastic search; mixture models;
D O I
10.1023/A:1008980332240
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
The Bayesian CART (classification and regression tree) approach proposed by Chipman, George and McCulloch (1998) entails putting a prior distribution on the set of all CART models and then using stochastic search to select a model. The main thrust of this paper is to propose a new class of hierarchical priors which enhance the potential of this Bayesian approach. These priors indicate a preference for smooth local mean structure, resulting in tree models which shrink predictions from adjacent terminal node towards each other. Past methods for tree shrinkage have searched for trees without shrinking, and applied shrinkage to the identified tree only after the search. By using hierarchical priors in the stochastic search, the proposed method searches for shrunk trees that fit well and improves the tree through shrinkage of predictions.
引用
收藏
页码:17 / 24
页数:8
相关论文
共 7 条
[1]  
Breiman L., 1984, BIOMETRICS, DOI DOI 10.2307/2530946
[2]   Bayesian CART model search [J].
Chipman, HA ;
George, EI ;
McCulloch, RE .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1998, 93 (443) :935-948
[3]  
Clark L. A., 1992, COMPSTAT 1990 P COMP
[4]  
Denison DGT, 1998, BIOMETRIKA, V85, P363
[5]  
HASTIE TJ, 1990, SHRINKING TREES
[6]   Monotone shrinkage of trees [J].
LeBlanc, M ;
Tibshirani, R .
JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS, 1998, 7 (04) :417-433
[7]  
Quinlan J. R., 1986, Machine Learning, V1, P81, DOI 10.1023/A:1022643204877