• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.55.2021.tde-19012022-161011
Document
Author
Full name
Fernanda Yuka Ueno
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2021
Supervisor
Committee
Santos, Maristela Oliveira dos (President)
Delbem, Alexandre Cláudio Botazzo
Munari Junior, Pedro Augusto
Nascimento, Mariá Cristina Vasconcelos
Title in Portuguese
Aprendizado de máquina em heurísticas de decomposição para problemas de dimensionamento de lotes
Keywords in Portuguese
Aprendizado de máquina
Heurística fix-and-optimize
Heurística relax-and-fix
Problema de dimensionamento de lotes
Abstract in Portuguese
Neste trabalho, são propostas heurísticas baseadas na partição do conjunto de variáveis dos modelos matemáticos, as quais são utilizadas para a resolução de dois problemas de dimensionamento de lotes. As heurísticas desenvolvidas são compostas de duas fases: construção de uma solução inicial e melhoria. As soluções iniciais são obtidas por meio da heurística relax-and-fix ou por um resolvedor comercial de otimização (primeira solução factível obtida). As heurísticas de melhoria são do tipo fix-and-optimize com partições clássicas e ADN (Automatically designed neighborhoods), que constrói uma vizinhança de forma automática, utilizando aprendizado de máquina não supervisionado, ou seja, usamos dois algoritmos de agrupamento: o k-means e o k-medoids. Nos experimentos computacionais, abordamos o problema de dimensionamento de lotes com múltiplas plantas distintas e o problema de dimensionamento de lotes multiestágio. Para realizar a comparação entre as heurísticas, foram utilizadas instâncias da literatura e as soluções são comparadas com as soluções obtidas por um otimizador comercial.
Title in English
Machine learning on decomposition heuristics for lot sizing problems
Keywords in English
Fix-and-optimize
Lot sizing problem
Machine learning
Relax-and-fix
Abstract in English
In this paper, heuristics based on the partition of the set of variables of the mathematical models are proposed, which are used to solve two lot sizing problems. The developed heuristics are composed of two phases: construction of an initial solution and improvement. The initial solutions are obtained through the relax-and-fix heuristic or by a commercial optimization solver (first feasible solution obtained). The improvement heuristics are of the fix-and-optimize type with classical partitions and ADN (Automatically designed neighborhoods), which builds a neighborhood automatically using unsupervised machine learning, i.e., we use two clustering algorithms, k-means and k-medoids. In the computational experiments we address the lot sizing problem with multiple distinct plants and the multistage lot sizing problem. To perform the comparison between the heuristics, instances from the literature were used and the solutions are compared with the solutions obtained by a commercial optimizer.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2022-01-19
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.