Algoritmo genético como instrumento para resolução, otimização e visualização do problema de localização de armazéns

• Ano de defesa: 2023
• Autor(a) principal: Milazzo, Flávio Oliveira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Elétrica
• Palavras-chave em Português: Logística; Transporte; Problema de localização de armazéns capacitados; Otimização através de algoritmos genéticos; Última milha; Comércio eletrônico; CNPQ::ENGENHARIAS; Engenharia elétrica; ODS::ODS 9. Indústria, Inovação e infraestrutura - Construir infraestrutura resiliente, promover a industrialização inclusiva e sustentável, e fomentar a inovação.
• Link de acesso: https://repositorio.ufu.br/handle/123456789/41060; http://doi.org/10.14393/ufu.di.2023.612

The problem of locating warehouses requires optimizing their locations. It is well known in operations research that it aims to minimize overhead costs and maximize customer satisfaction in supply chain management. Location solutions require the use of mathematical models, algorithms and techniques that can be applied, depending on the complexity of the problem and the related data situation (brute force, linear programming, integer programming and heuristics, for example). Electronic commerce or e-commerce has grown significantly over the last few decades and product transport companies are concerned with facilitating logistics operations. The subject is the challenges posed by locating capable warehouses in relation to logistics demand. The aim of this study is to define a set of warehouse locations and connect them to the customer contingent, minimizing total opening, closing and/or maintenance costs, preserving customer satisfaction (added value) and reducing or eradicating the capacity constraints of these DCs. The hypothesis is that the distances between storage locations and the arrival of transport vehicles require an efficient and optimized operating system that overcomes the models and shortcomings offered by traditional logistics. The question is how to solve the problem of locating capable warehouses in a company's supply chain in order to ensure better quality. the solution could be the application of the Genetic Algorithm. Objective: to present a Genetic Algorithm with some restrictions that is capable of promoting the optimization and on-demand visualization of solutions or approximations to the problem of locating capable warehouses. Method: presentation of the Algorithm model developed by the author, based on the network of participating suppliers who have applied to lease warehouses in the same environmental space or close to these applicants. Database - Kaggle database, by Antony Goldbloom, with the aim of hosting Data Science competitions and providing data on various topics (datasets). We used the dataset "E-Commerce Brazilian Data Set with 100k Orders from 2016 to 2018 - Olist (a public Brazilian e Commerce dataset, provided by Olist Store in the marketplace segment. Optimal Solution - Mixed Integer Linear Programming (MILP) which seeks the optimal solution to the problem of locating capable warehouses; an open source Python library called PULP was used. Genetic algorithm: the Distributed Evolutionary Algorithms (DEAP) library was used. Conclusion: the results of the algorithm presented are promising. In proportion to the number of generations, there was an advantageous evolution of the fitness function and the consequent approximation to the optimal solution calculated by MILP. seeking the optimal solution to the problem of locating capable warehouses, an open source Python library called PULP was used. The Distributed Evolutionary Algorithms (DEAP) library was used for the Genetic Algorithm. It was concluded that the results of the algorithm presented are promising. In proportion to the number of generations, there was an advantageous evolution of the fitness function and the consequent approximation to the optimal solution calculated by MILP.