The choice of the optimal amount of food to load on each individual train is a crucial topic for the onboard high-speed railway catering service. In fact, a correct management of food load implies important goals: the decreasing of food waste, the reduction of non-value-added activities, it means non-necessary food load and the consequent unloading, and the optimization in the use of warehouses, on board and on stations. This paper applies a methodology for automating the entire process of load forecast of food on trains using classification algorithms of machine learning. Starting from a collection of data about travel time, route, duration and other external factors, the goal is to predict the optimal load of every type of product for the specific train. A real case study positively validates the load forecast methodology, using a 4 months database of 14 parameters from some trains circulating throughout Italy. The paper shows the results provided by the comparison of four supervised classification algorithms, the logistic regression (LR), the neural network (NN), the decision forests (DF) and the decision jungles (DJ). We have compared the algorithms in two situations: on the product that has been wasted more frequently, and on all non-seasonal products. The results suggest preferring LR for one product classification and NN for multi-product classification.

Load forecast for onboard high-speed railway catering service using machine learning / Quatrini, E.; Costantino, F.; Di Gravio, G.; Patriarca, R.. - In: ...SUMMER SCHOOL FRANCESCO TURCO. PROCEEDINGS. - ISSN 2283-8996. - 1:(2019), pp. 382-388. ((Intervento presentato al convegno XXIV Summer School “Francesco Turco” – Industrial Systems Engineering tenutosi a Brescia; Italy.

Load forecast for onboard high-speed railway catering service using machine learning

Quatrini E.
;
Costantino F.;Di Gravio G.;Patriarca R.
2019

Abstract

The choice of the optimal amount of food to load on each individual train is a crucial topic for the onboard high-speed railway catering service. In fact, a correct management of food load implies important goals: the decreasing of food waste, the reduction of non-value-added activities, it means non-necessary food load and the consequent unloading, and the optimization in the use of warehouses, on board and on stations. This paper applies a methodology for automating the entire process of load forecast of food on trains using classification algorithms of machine learning. Starting from a collection of data about travel time, route, duration and other external factors, the goal is to predict the optimal load of every type of product for the specific train. A real case study positively validates the load forecast methodology, using a 4 months database of 14 parameters from some trains circulating throughout Italy. The paper shows the results provided by the comparison of four supervised classification algorithms, the logistic regression (LR), the neural network (NN), the decision forests (DF) and the decision jungles (DJ). We have compared the algorithms in two situations: on the product that has been wasted more frequently, and on all non-seasonal products. The results suggest preferring LR for one product classification and NN for multi-product classification.
File allegati a questo prodotto
File Dimensione Formato  
Quatrini_Load_2019.pdf

solo gestori archivio

Tipologia: Documento in Post-print (versione successiva alla peer review e accettata per la pubblicazione)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 402.78 kB
Formato Adobe PDF
402.78 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11573/1346942
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact