Título: A Machine Learning approach to predict Length of Stay of vehicles in an inbound logistics operation
Autores: Victor Hugo Soares Pereira, Kaike Sa Teles Rocha Alves, Eduardo Pestana de Aguiar
Resumo: A large steel plant receives up to millions of tons of scrap metal through road transportation each year in its inbound logistics process. The Length of Stay (LOS) of vehicles is one of the most critical metrics that represent the performance of the unloading operation of raw materials. Accurately predicting this metric enables managers to make data-driven decisions at operational, tactical, and strategic levels. This study proposes implementing a Machine Learning (ML) approach for predicting the LOS of vehicles loaded with scrap metal in the inbound operation of a large Brazilian steel plant. The performance of five ML models – Linear Model Ridge, k-nearest neighbors Regressor, Gradient Boosting Regressor, Decision Tree Regressor, and ePL-KRLS-DISCO – was evaluated in terms of Root Mean Squared Error (RMSE), Mean Average Error (MAE), and execution time. The results are compared with the current method of prediction and statically validated through an Analysis of Variance (ANOVA) test. The ML approach applied in this study achieved better accuracy, reducing by 64% the RMSE, and has the potential to enable more reliable data-driven decisions for the company
Palavras-chave: Machine Learning, Steel Industry, Logistics, Regression models
Páginas: 6
Código DOI: 10.21528/CBIC2023-005
Artigo em pdf: CBIC_2023_paper005.pdf
Arquivo BibTeX: CBIC_2023_005.bib