A probabilistic cellular automaton to forecast port choice decisions

Abstract: The port choice problem consists in predicting the selection of a port, made by an agent who has alternatives to choose from. Most of the literature has tackled this problem assuming a discrete choice model dependent on the ports characteristics and agents attributes. However, in practice the port choice decision depends also on the choices made by other agents as well as decisions made by these agents in the past. There are only a few examples that incorporate the complexity generated by spatio-temporal interactions between agents. However, those modelling structures are rather cumbersome, precluding their use in practical cases. This article presents a new modelling framework to predict port choice decisions, based on the theory of Cellular Automaton (CA), which is simple in structure and can be quickly calibrated and applied. This framework is a probabilistic CA intended to imitate the decision processes made from multiple shippers that interact with each other. These shippers face similar alternatives of seaports for exporting their products within a certain time span. The port choice here is a dynamic decision that depends on the ports characteristics and attributes of each shipper at a given time, as well as the decisions made by their neighbours. The outcome of the interaction is a discrete decision that evolves in time according to the dynamics of the system as a whole. The specified CA was applied to the case of vehicle exports from Brazil and the calibration was performed through a genetic algorithm. The results show that the probabilistic CA is able to replicate the historic behaviour of the port choice decisions in the Brazilian vehicle industry, with a high degree of success. The spatial component of the CA turned out to be of major relevance in the dynamic decision process along with the attributes and geographical location of ports.

 Fuente: Transport, 2018, 33(3): 801-809

Editorial: Taylor & Francis

 Año de publicación: 2018

Nº de páginas: 9

Tipo de publicación: Artículo de Revista

DOI: 10.3846/transport.2018.5478

ISSN: 1648-4142,1648-3480

Url de la publicación: https://doi.org/10.3846/transport.2018.5478