Webinar video https://www.youtube.com/watch?v=gWj8T69k6wg
Abstract
Machine Learning provides powerful predictive tools, but many models are seen as black boxes, yielding predictions that are difficult to interpret and explain. Generalized Linear Models (GLMs), in contrast, offer interpretable approximations and lead to much more tractable optimization problems. In this talk, we will illustrate the successful use of GLMs in two key challenges in classification and regression: feature selection at both local and global levels, and the design of counterfactual decisions, i.e., determining the minimal perturbation to a record or set of records needed to achieve a desired prediction.
Short Bio
Emilio Carrizosa is Professor of Statistics and Operations Research at the University of Seville, and President of math-in, the Spanish Network for Mathematics and Industry. He has also served as President of SEIO, the Spanish Society of Statistics, Operations Research and Data Science. He has published over 150 papers in international journals, primarily in Operations Research, but also in Statistics and applied fields such as Energy, Chemical Engineering and Hydrology. His work has been recognized with awards such as the SEIO Medal 2024 and the SEIO-FBBVA Award 2024 for the best paper in Statistics and Operations Research applied to Big Data and Data Science.
Carrizosa is also involved in knowledge transfer and outreach activities, frequently discussing in media on industrial mathematics and mathematics education.
Information at FEP News
More information at News U. Porto