Activity: Academic talk or presentation types › Oral presentation
Description
We propose a data-driven robust optimization approach where cluster Voronois are used to identify and discard anomalous regions of the uncertainty set. We identify anomalous regions via sparse clusters and construct Voronois using perpendicular bisecting hyperplanes. With the recognition that not all anomalous data result in anomalous decisions, we also develop a method to maximize the size of the non-anomalous regions, such that decisions remain non-anomalous. Our anomaly-based models show marked improvements in performances over the classical robust optimization with polyhedral uncertainty on a disaster response model that uses real data from the last 18 years of impacts of floods and landslides in Brazil.