ECAS – SFdS course

Random forests: basics, extensions and applications

October 8-13, 2023 – La Villa Clythia – CAES du CNRS, Fréjus, France

Main topic

Random forests are one of the state-of-the-art methods in machine learning. They are often presented as one of the best « off-the-shelf » prediction methods both for regression and supervised classification problems, mainly because they give competitive results with only a few hyper-parameters to tune. Since their introduction by Leo Breiman in 2001, many extensions have been proposed to solve different prediction problems (e.g., time series, survival data, longitudinal data).

In this course, starting from the basics on tree-based methods and popular tree ensemble methods, we will present several extensions of random forests and demonstrate their ability in applications coming from several domains. We will also discuss the link between random forests and other popular machine learning techniques, and give some clues about the use of some more abstract tools to analyse the behavior of random forests.

Some experience in programming with R and/or Python is a plus but not a prerequisite. No prior knowledge on tree-based methods is required.

Central illustration credit: Marc Varachaud, original creation

©2024 SFdS
Société Française de Statistique
Institut Henri Poincaré
11 rue Pierre et Marie Curie
75231 Paris cedex 5
Tél. : +33 (0)1 44 27 66 60
Notre site a été supporté par :