JSM 2014 Home
Online Program Home
My Program

Abstract Details

Activity Number: 331
Type: Contributed
Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
Sponsor: Section on Nonparametric Statistics
Abstract #313258 View Presentation
Title: Detecting Interactions in Supervised Ensemble Learning Algorithms
Author(s): Lucas Mentch*+ and Giles Hooker
Companies: Cornell University and Cornell University
Keywords: Interactions ; Bagging ; Random Forests ; CART ; Trees ; Statistical Inference
Abstract:

Ensemble methods in machine learning, such as bagging and random forests, remain among the most popular supervised learning schemes due to their flexibility and minimal required tuning. However, formal statistical inference methods have largely remained absent from learning contexts due to the inherent complexity of the algorithms. Recently, we showed that predictions generated by bagging and random forest algorithms are asymptotically normal whenever proper subsamples of the training set are used to build each tree in the ensemble. Here, we extend this previous work to produce a formal hypothesis test for detecting interactions between features. Since tests for interactions exist in other more traditional contexts like linear regression, we also compare the power of our testing procedure to that of the more established procedures that accompany these simpler, more interpretable models. Both theoretical results and simulations are provided.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2014 program




2014 JSM Online Program Home

For information, contact jsm@amstat.org or phone (888) 231-3473.

If you have questions about the Professional Development program, please contact the Education Department.

The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.

ASA Meetings Department  •  732 North Washington Street, Alexandria, VA 22314  •  (703) 684-1221  •  meetings@amstat.org
Copyright © American Statistical Association.