Ships arrival times prediction using deep learning

The objective of this research project is to use Deep Learning in predicting ships’ arrival times of a medium size multi-commodity bulk port. The outcome of this project will serve dynamically in updating berths schedules to minimize ports turnaround time and build models closer to reality. We will be using ships trajectories data or AIS data combined with some exogenous data such as weather data, physical structure of ships, Machine conditions, unforeseen events and seasonality. Also, we will focus more on the real time estimation of arrival times by building flexible model for short-term and long-term prediction horizons. A first objective will be to prepare properly the data that will be used in building the predictive model, as the problem may arise in real-time data management in terms of computing and storage resources. Also, the use of multiple data sources will require a specific treatment related to data handling and feature engineering. The second objective will be to build models based on Artificial Neural Networks, train them and select the best one. Then, the last objective will be to deploy the chosen model in order to make prediction on production data.

Faculty Supervisor:

Yoshua Bengio;Loubna Benabbou

Student:

Partner:

École Mohammadia d'Ingénieurs, Mohammed V University in Rabat

Discipline:

Engineering

Sector:

Education

University:

Université de Montréal

Program: