Regression with missing data, a comparison study of techniques based on random forests

Abstract

In this paper, we present the practical benefits of a new random forest algorithm to deal with missing values in the sample. The purpose of this work is to compare the different solutions to deal with missing values using random forests and describe the new algorithm performance as well as its algorithmic complexity. A variety of data-missing mechanisms (MCAR, MAR, MNAR) are considered and simulated. We study the quadratic errors and the bias of our algorithm and compare them to the most popular missing values random forests algorithms in the literature. In particular, we compare those techniques for both regression and prediction purposes. This work follows the paper of Gómez-Méndez and Joly [On the consistency of a random forest algorithm in the presence of missing entries. 2020. Available from: arXiv:201105433] on the consistency of this new algorithm.