Rolling Lookahead Learning for Optimal Classification Trees
From MaRDI portal
Publication:6433857
arXiv2304.10830MaRDI QIDQ6433857
Enis Kayış, Taghi Khaniyev, Zeynel Batuhan Organ
Publication date: 21 April 2023
Abstract: Classification trees continue to be widely adopted in machine learning applications due to their inherently interpretable nature and scalability. We propose a rolling subtree lookahead algorithm that combines the relative scalability of the myopic approaches with the foresight of the optimal approaches in constructing trees. The limited foresight embedded in our algorithm mitigates the learning pathology observed in optimal approaches. At the heart of our algorithm lies a novel two-depth optimal binary classification tree formulation flexible to handle any loss function. We show that the feasible region of this formulation is an integral polyhedron, yielding the LP relaxation solution optimal. Through extensive computational analyses, we demonstrate that our approach outperforms optimal and myopic approaches in 808 out of 1330 problem instances, improving the out-of-sample accuracy by up to 23.6% and 14.4%, respectively.
Has companion code repository: https://github.com/koftezz/rolling-lookahead-DT
This page was built for publication: Rolling Lookahead Learning for Optimal Classification Trees
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6433857)
