Optimal decision trees for categorical data via integer programming

被引:36
作者
Gunluk, Oktay [4 ]
Kalagnanam, Jayant [1 ]
Li, Minhan [2 ]
Menickelly, Matt [3 ]
Scheinberg, Katya [4 ]
机构
[1] IBM Res, Yorktown Hts, NY USA
[2] Lehigh Univ, Bethlehem, PA 18015 USA
[3] Argonne Natl Lab, Lemont, IL USA
[4] Cornell Univ, Ithaca, NY 14850 USA
关键词
Decision trees; Integer programming; Machine learning; Binary classification;
D O I
10.1007/s10898-021-01009-y
中图分类号
C93 [管理学]; O22 [运筹学];
学科分类号
070105 ; 12 ; 1201 ; 1202 ; 120202 ;
摘要
Decision trees have been a very popular class of predictive models for decades due to their interpretability and good performance on categorical features. However, they are not always robust and tend to overfit the data. Additionally, if allowed to grow large, they lose interpretability. In this paper, we present a mixed integer programming formulation to construct optimal decision trees of a prespecified size. We take the special structure of categorical features into account and allow combinatorial decisions (based on subsets of values of features) at each node. Our approach can also handle numerical features via thresholding. We show that very good accuracy can be achieved with small trees using moderately-sized training sets. The optimization problems we solve are tractable with modern solvers.
引用
收藏
页码:233 / 260
页数:28
相关论文
共 17 条
[1]  
[Anonymous], Balance
[2]  
Bennett KP, 1998, IEEE WORLD CONGRESS ON COMPUTATIONAL INTELLIGENCE, P2396, DOI 10.1109/IJCNN.1998.687237
[3]   Classification and regression via integer optimization [J].
Bertsimas, Dimitris ;
Shioda, Romy .
OPERATIONS RESEARCH, 2007, 55 (02) :252-271
[4]   Optimal classification trees [J].
Bertsimas, Dimitris ;
Dunn, Jack .
MACHINE LEARNING, 2017, 106 (07) :1039-1082
[5]   Random forests [J].
Breiman, L .
MACHINE LEARNING, 2001, 45 (01) :5-32
[6]   LIBSVM: A Library for Support Vector Machines [J].
Chang, Chih-Chung ;
Lin, Chih-Jen .
ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY, 2011, 2 (03)
[7]  
Dash S, 2018, ADV NEUR IN, V31
[8]  
FICO, 2018, EXPL MACH LEARN CHAL
[9]  
Hyafil L., 1976, Information Processing Letters, V5, P15, DOI 10.1016/0020-0190(76)90095-8
[10]   Decision trees: a recent overview [J].
Kotsiantis, S. B. .
ARTIFICIAL INTELLIGENCE REVIEW, 2013, 39 (04) :261-283