Data-driven Prediction of Relevant Scenarios for Robust Combinatorial Optimization

TL;DR

This paper introduces a machine learning heuristic to select effective initial scenarios in robust combinatorial optimization, enhancing solution quality and providing insights into scenario importance.

Contribution

It presents a novel, dimension-independent feature-based Random Forest approach for scenario selection in robust optimization, scalable to larger instances.

Findings

Improves solution bounds for larger problem instances

Provides feature importance insights into scenario properties

Demonstrates effectiveness of ML heuristic across various sizes

Abstract

We study iterative methods for (two-stage) robust combinatorial optimization problems with discrete uncertainty. We propose a machine-learning-based heuristic to determine starting scenarios that provide strong lower bounds. To this end, we design dimension-independent features and train a Random Forest Classifier on small-dimensional instances. Experiments show that our method improves the solution process for larger instances than contained in the training set and also provides a feature importance-score which gives insights into the role of scenario properties.