Targeting predictors in random forest regression

TL;DR

This paper introduces a targeting method for random forest regression that improves predictive accuracy by controlling split placement along strong predictors, especially in high-dimensional sparse data settings.

Contribution

It proposes a novel targeting approach that enhances split selection in RF, balancing bias and variance, and demonstrates significant accuracy improvements in economic data applications.

Findings

Targeting increases individual tree strength.

Optimal targeting uses 10-30% of predictors.

Predictive accuracy improves by up to 13%.

Abstract

Random forest regression (RF) is an extremely popular tool for the analysis of high-dimensional data. Nonetheless, its benefits may be lessened in sparse settings due to weak predictors, and a pre-estimation dimension reduction (targeting) step is required. We show that proper targeting controls the probability of placing splits along strong predictors, thus providing an important complement to RF's feature sampling. This is supported by simulations using representative finite samples. Moreover, we quantify the immediate gain from targeting in terms of increased strength of individual trees. Macroeconomic and financial applications show that the bias-variance trade-off implied by targeting, due to increased correlation among trees in the forest, is balanced at a medium degree of targeting, selecting the best 10--30\% of commonly applied predictors. Improvements in predictive accuracy of targeted RF relative to ordinary RF are considerable, up to 12-13\%, occurring both in recessions and expansions, particularly at long horizons.