Título: A Machine Learning Application in Brazilian Railway Crew Rostering
Autores: Rael Fonseca Andretto, Eduardo Pestana de Aguiar
Resumo: A large portion of railway expenses is in the workforces payment. Although there are specialized professionals whose function is to appoint the drivers to the trains economically, many uncertainties exist at the time of the decision, leading many drivers to carry out idle cycles. Machine learning techniques are ideal for generalizing new scenarios from a training database and can be applied to reduce uncertainties in many problems. The literature shows that there are few machine learning applications in crew scheduling and rostering, especially in railways. In this study, applied at a Brazilian railway operator, in the first step, exploratory data analysis techniques are used together with the rules generated by a decision tree to create and apply guidelines to reduce idle crew cycles. In the second step, five machine learning algorithms are evaluated to automate and improve the process: neural network, support vector machine, decision tree, random forest, and the Autonomous Learning Multi-Model System (ALMMo). Although the first step got acceptable results and has been applied in the company, the machine learning models improved the result, showing an accuracy above 86% on average, which meets all service levels established by the company. Finally, the decision tree, the random forest, and the ALMMo were considered suitable solutions for application in the company due to their performances and characteristics
Palavras-chave: railway, machine learning, crew scheduling
Páginas: 7
Código DOI: 10.21528/CBIC2023-023
Artigo em pdf: CBIC_2023_paper023.pdf
Arquivo BibTeX: CBIC_2023_023.bib