Description
We propose a data-driven robust optimization approach where cluster Voronois are used to identify (via sparsity) and discard anomalous regions of the uncertainty set. The Voronois are built using perpendicular bisecting hyperplanes and form a geometrical characterisation of data clusters. Combined with a geometrical formulation of the uncertainty support, our data-driven uncertainty set is equipped with a comprehensive geometrical intuition, with enough generality to allow decision makers to tailor supports based on the problem at hand. 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.| Period | 13 Sept 2022 → 15 Sept 2022 |
|---|---|
| Event title | OR64 Operational Research Society Annual Conference |
| Event type | Conference |