ETA prediction for containerships at the Port of Rotterdam using Machine Learning Techniques

ETA prediction for containerships at the Port of Rotterdam using Machine Learning Techniques

Parolas, I.

van Duin, J.H.R. (mentor)

Technology, Policy and Management

Management of Technology

2016-08-24

The hinterland transportation of incoming containers at container terminals is a complex problem, due to the various actors involved and their often conflicting interests. A promising solution towards the problem for hinterland network operators is that of synchromodality, a concept that refers to on-line network planning for hinterland transportation. However, a hindrance to the efficient planning and execution of hinterland transportation is that there is currently no accurate way of predicting the estimated time of arrivals (ETA) for containerships that are reaching container terminals. This results in huge uncertainty over the types and amounts of cargo that reach the terminals, which in turn hinders the fast and cost efficient distribution of the products to inland destinations through trucks, trains or barges. The current paper will propose a machine learning approach for predicting the ETA of containerships heading towards the Port of Rotterdam, by combining position data from GPS signals with weather predictions. It was found that significant improvement for the ETA predictions, compared to the current situation could be achieved, especially for the cases of the vessels that are more than 60 hours away from the port. Furthermore, the weather interpretation was not of significant importance for estimating the time of vessel arrivals at the port. The value of such an information tool for the various stakeholders involved was also investigated. The interested parties, for which the importance of ETA predictions of sea vessels was assessed are : terminal operators (European Container Terminals in the case at hand), hinterland transportation companies (e.g. European Gateway Services ), the Port of Rotterdam, carriers and importers.

machine learning
neural networks
support vector machines
estimated time of arrival
container transport
hinterland transportation

http://resolver.tudelft.nl/uuid:9e95d11f-35ba-4a12-8b34-d137c0a4261d

Student theses

master thesis

(c) 2016 Parolas, I.