Tree-Based Machine Learning Methods Workshop

COURSE OVERVIEW

This half-day virtual workshop gives a practical, code-centered introduction to tree-based machine learning methods using R. Tree-based methods are useful for nonlinear signals, mixed data types, robustness, and scalable prediction. The workshop focuses on random forest ensembles and centers on the R-package randomForestSRC, which implements random forests for regression, classification, and survival. Topics include out-of-bag inference, test-data prediction, and variable selection using permutation VIMP, minimal depth, and rule-based variable priority with varPro. Advanced topics cover class imbalance, imputation and test time OOD, random hazard forests (RHF), and super greedy trees (SGT). Related packages in the randomForestSRC ecosystem will also be discussed.

WORKSHOP OUTLINE

1. Training

• Regression, classification, survival
• Examples and R code
• Worked interpretations

2. Inference and Prediction

• Out-of-bag inference
• Test-data prediction
• Restore mode and partial plots

3. Variable Selection

• Permutation VIMP
• Minimal depth
• Variable priority with varPro

4. Advanced Topics

• Class imbalance
• Imputation and test time OOD
• Random Hazard Forests
• Super Greedy Trees

SOFTWARE ECOSYSTEM

Designed for applied statisticians and data scientists who want hands-on R workflows for prediction, inference, and variable selection.

• randomForestSRC: Unified forests for regression, classification, survival, and competing risks.
• varPro: Rule-based variable priority for model-independent variable selection.
• randomForestRHF (RHF): Random hazard forests for time-varying covariates in survival analysis.
• randomForestSGT (SGT): Super greedy trees and flexible split geometry.

INSTRUCTORS

Workshop leaders for Tree-Based Machine Learning Methods

Hemant Ishwaran

• Professor of Public Health Sciences
• Graduate Program Director; Director of Statistical Methodology
• Division of Biostatistics, University of Miami

Hemant Ishwaran develops machine learning methods for complex biomedical and time-to-event data and turns them into practical open-source tools for investigators. He created Random Survival Forests and the R package randomForestSRC. His work has been applied in cardiovascular disease, heart transplantation, cancer, and genomics.

https://ishwaran.org/ishwaran.html

Min Lu

• Research Associate Professor
• Division of Biostatistics, University of Miami

Min Lu works on random forests and trees, causal inference, variable selection, infectious disease modeling, statistical genomics, and meta-analysis. She develops methods and software in the randomForestSRC and varPro ecosystem and collaborates on practical applications in medicine and public health.

https://www.luminwin.net/

PLEASE NOTE:

• Registrants will be able to access the workshop recording for three weeks following the event.
• All ticket sales are considered final. We do not offer refunds.