A prediction model for maritime freight container rates

Abstract

iContainers is an on-line maritime freight forwarder firm that offers real time prices on its website for container freight transportation services. This business model has allowed to keep a record of historical rates, although until now it has been never used with analysis purposes. This master thesis aims to change that, using rates data in order to build models with the ultimate objective of forecasting. It covers all the stages in a typical data-based project: data collection and reliability check, data exploration, modelling, forecasting and finally implementation of the results. Transportation rates data are not trivial, since they have both a time and a space dimension. In this context, statistical techniques that are able to deal with both have been used. Amongst them, it has been of utmost importance the use of multinomial bayesian generalised mixed models which, inspired by Markov chains, act as transition models that allow to infer how likely it is for a rate to go up, down or remain stable in the future.