Matthew Baucum
Matthew Baucum
Assistant Prof. of Business Analytics at Florida State University College of Business
Matt Baucum is an Assistant Professor of Business Analytics, Information Systems, and Supply Chain at Florida State University. His research focuses on the use of machine and reinforcement learning to address healthcare and policy challenges, and his primary teaching interests are data visualization and data mining. Prior to his academic career, Dr. Baucum worked in market research at Honda Research & Development and in user experience research at AT&T. He has presented at 10 analytics-focused academic conferences, and has published in high-quality data analytics outlets such as Management Science and the INFORMS Journal on Data Science.
Track: Marketing
We Can Stop Saying “Black Box” Now: Extracting Clear, Actionable Insights from High-Dimensional Machine Learning Models
Machine learning (ML) classifiers and regressors are often thought of as ‘black box’ models, offering strong predictive performance at the expense of interpretability. This presentation pushes back against this narrative by surveying existing and novel ML techniques for discovering clear patterns in high-dimensional business datasets. The presentation will provide an easy-to-code, easy-to-interpret framework for discovering (1) predictors with the greatest impact on an outcome of interest, (2) predictors with highly nonlinear effects, and (3) interactions between predictors. The presentation will include a novel visualization method for decomposing each predictor’s effect into its linear, nonlinear, and interaction component, and a case study exemplifying the practical value of the presented approaches.
