Port sustainability management based on a bayesian network model application to the spanish port system

Abstract

Numerous variables are involved in port operation and management, and infrastructure managers need to know relationships between them, to be able to modify operating conditions. Using Bayesian networks makes possible to classify, predict and diagnose these variables, allowing to estimate the posterior probability of the unknown variables based on the known variables. This means, at the planning level, that it is not necessary to know all the variables, knowing their relationships. Therefore, these networks can be used to make optimal decisions introducing possible actions and usefulness of their results. In the methodology proposed, a database with more than 100 port variables has been generated. Variables are classified as economic, social, environmental and institutional, as it is made in smart port studies in all Spanish Port System. Using this database, a network has been generated using an acyclic directed graph, which allows knowing relations between port variables regarding parents and sons. This kind of network allows modeling uncertainty probabilistically even when the number of variables is high as in the case of port planning and exploitation. The main conclusion of the study is that economic variables are the cause of the other typologies and they play the role of parents in the network in most of the cases. Moreover, another conclusion is related to environmental variables. So, if they are known, the network will allow estimating the subsequent probability of social ones.