SC03: Modern Machine Learning with Bayesian Additive Regression Trees (BART) (Additional Fee)

Relatively cheap, off-the-shelf computing power has led to breakthroughs in our ability to learn high-dimensional, complex relationships from large, curated data sets. The two key modeling approaches in this area are deep learning with neural networks and ensemble learning with trees. Deep learning is an ideal choice for prediction when all the covariates are of the same type (e.g., pixels, words, audio waves). Conversely, ensemble learning is superior with respect to out-of-sample predictive performance for tabular data when all the covariates are of different types (e.g., age, sex, weight). A collection of machines (trees in this case) is fit simultaneously and forms the basis of an ensemble's aggregate predictive performance far beyond that of any single machine's fit. The focus of this workshop is machine learning with an ensemble of trees with Bayesian Additive Regression Trees. Although its machine learning foundation might seem daunting, BART is easy to use. We will demonstrate with user-friendly R packages and worked examples. The Bayesian approach allows for a Markov chain Monte Carlo stochastic exploration of the model space, uncertainty quantification, and posterior inference. BART is a modern nonparametric approach that exploits the elegance and convenience of the Bayesian paradigm. Handling outcomes of different types—continuous, dichotomous/categorical, and time-to-event—is one of BART's strengths. Furthermore, we will demonstrate BART's effectiveness in a wide range of regression applications, including marginal effects, variable selection, monotonicity, outlier detection, and time-to-event extensions such as competing risks or recurrent events. This is a lot of material to cover in one day, so we have developed digital materials for attendees to self-explore afterward.