Flowshop NEH-Based Heuristic Recommendation
| Autor: | Marie-Eléonore Kessaci, Myriam Regattieri Delgado, Lucas M. Pavelski |
|---|---|
| Přispěvatelé: | Federal University of Technology - Paraná (UTFPR), Operational Research, Knowledge And Data (ORKAD), Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 (CRIStAL), Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS)-Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS), Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS) |
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
Algorithm configuration
Mathematical optimization 021103 operations research Heuristic Computer science 0211 other engineering and technologies Decision tree 02 engineering and technology [INFO.INFO-RO]Computer Science [cs]/Operations Research [cs.RO] [INFO.INFO-DM]Computer Science [cs]/Discrete Mathematics [cs.DM] Random forest Set (abstract data type) Test set 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Heuristics ComputingMilieux_MISCELLANEOUS |
| Zdroj: | European Conference on Evolutionary Computation in Combinatorial Optimization (EvoCOP) European Conference on Evolutionary Computation in Combinatorial Optimization (EvoCOP), Apr 2021, Seville, Spain. pp.136-151, ⟨10.1007/978-3-030-72904-2_9⟩ Evolutionary Computation in Combinatorial Optimization ISBN: 9783030729035 EvoCOP |
| Popis: | Flowshop problems (FSPs) have many variants and a broad set of heuristics proposed to solve them. Choosing the best heuristic and its parameters for a given FSP instance can be very challenging for practitioners. Per-instance Algorithm Configuration (PIAC) approaches aim at recommending the best algorithm configuration for a particular instance problem. This paper presents a PIAC methodology for building models to automatically configure the Nawaz, Encore, and Ham (NEH) algorithm which proved to be a good choice in most FSP variants (especially when they are used to provide initial solutions). We use irace to build the performance dataset (problem features \(\leftrightarrow \) algorithm configuration), while training Decision Tree and Random Forest models to recommend NEH configurations on unseen problems of the test set. Results show that the recommended heuristics have good performance, especially those by random forest models considering parameter dependencies. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|