A Simple Experiment to Test the Interaction Effects Hypothesis for Random Forests and Logistic Regression

7 Pages Posted: 22 Dec 2011 Last revised: 2 Jan 2012

Date Written: December 22, 2011

Abstract

This article discusses how to build logistic regression models which take interactions into effect. The paper discusses the interaction effects hypothesis for ensemble methods which states random forests and ensemble methods dominate logistic regression because they optimal detect and use predictive interaction effects in the data. This hypothesis is supported using a simple intuitive synthetic data experiment. The importance of this fact is that when random forests outperform logistic regression it means there are predictive interaction effects in the variables which can be added to the logistic regression and when random forests do not outperform there are no predictive interaction effects in the data. This is extremely useful for predictive modeling and statistical analysis.

Keywords: random forest, interaction effects, logistic regression

Suggested Citation

Sharma, Dhruv, A Simple Experiment to Test the Interaction Effects Hypothesis for Random Forests and Logistic Regression (December 22, 2011). Available at SSRN: https://ssrn.com/abstract=1975798 or http://dx.doi.org/10.2139/ssrn.1975798

Dhruv Sharma (Contact Author)

Independent ( email )

2023 N. Cleveland St.
Arlington, VA 22201
United States

HOME PAGE: http://theinterdisciplinarian.com/

Here is the Coronavirus
related research on SSRN

Paper statistics

Downloads
188
Abstract Views
1,132
rank
177,530
PlumX Metrics