EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2771. Applying Machine Learning in Machine Scheduling - Improving the Decision-Making of a Serial-Batch Scheduling Problem
Invited abstract in session WC-60: Machine Learning in Machine Scheduling, stream Project Management and Scheduling.
Wednesday, 12:30-14:00Room: S09 (building: 101)
Authors (first author is the speaker)
| 1. | Aykut Uzunoglu
|
| Chair of Production & Supply Chain Management, Augsburg University | |
| 2. | Christian Gahm
|
| Chair of Production & Supply Chain Management, Augsburg University | |
| 3. | Axel Tuma
|
| Faculty of Business Administration , University of Augsburg |
Abstract
Heuristic algorithms are commonly used in practical applications of combinatorial optimization to achieve efficient problem-solving. However, these heuristics have limitations when applied to interleaved and computationally hard problems. To address these shortcomings, we propose enhancing decision-making by using Machine Learning (ML) models. We demonstrate the effectiveness of this approach on a complex example of a serial-batch scheduling problem, which involves deciding the grouping of jobs (i.e. batching) and sequencing of batches (i.e. scheduling). The maximum potential of ML models can be exploited by strategically applying them at various junctures within the problem-solving process. We analyze three stages to integrate ML models to improve decision-making. Firstly, ML models are used to anticipate solutions for base-level nesting problems posed by the batching decisions. Secondly, the models are used to tune the parameters of the top-level problem's solution method circumventing time-intensive search approaches. Finally, the most suitable solution method is selected for a problem instance by creating ordinal preferences between solution methods based on model predictions. The effectiveness of using ML models to improve decision-making at different stages of the scheduling problem is highlighted by our empirical evaluations.
Keywords
- Combinatorial Optimization
- Machine Learning
- Scheduling
Status: accepted
Back to the list of papers