590. Machine learning meets optimization: Edge classification for routing problems
Invited abstract in session TD-58: Machine Learning and Artificial Intelligence, stream Vehicle Routing and Logistics.
Tuesday, 14:30-16:00Room: Liberty 1.13
Authors (first author is the speaker)
| 1. | Stefan Voigt
|
| Department of Supply Chain Management & Operations, Catholic University of Eichstätt-Ingolstadt | |
| 2. | Johannes Gückel
|
| Department of Operations, Katholische Universität Eichstätt-Ingolstadt | |
| 3. | Pirmin Fontaine
|
| Ingolstadt School of Management, Catholic University of Eichstätt-Ingolstadt |
Abstract
Routing problems, such as the Traveling Salesman Problem (TSP) and the Capacitated Vehicle Routing Problem (CVRP), pose significant computational challenges due to their large search spaces. Traditional optimization methods, including exact solvers and (meta)heuristics, often struggle with efficiency as the problem size grows. In this study, we propose a novel machine learning-based approach to classify edges within routing problems, predicting the likelihood of each edge being part of an optimal solution. By using supervised learning techniques, we systematically reduce the search space, enhancing both solver performance and heuristic efficiency. We evaluate multiple machine learning models, analyzing their predictive accuracy and impact on solution quality of solvers and state-of-the-art metaheuristics. Our results demonstrate that incorporating machine learning-based preprocessing significantly improves computational efficiency without compromising solution optimality. Additionally, we provide insights into the key factors influencing edge selection and discuss implications for real-world logistics and transportation applications.
Keywords
- Vehicle Routing
- Machine Learning
- Metaheuristics
Status: accepted
Back to the list of papers