Deep Learning Forecasting for Supporting Terminal Operators in Port Business Development

Accurate forecasts of containerised freight volumes are unquestionably important for port terminal operators to organise operations and develop business plans. They also relevant authorities, regulators, governmental agencies dealing with transportation. In a time when deep learning is in the limelight, owing consistent strip success stories, it natural apply tasks forecasting container throughput. Given number options, practitioners can benefit from lessons learned applying models problem. Coherently, this work, we devise multivariate predictive based on learning, analysing assessing their performance identify architecture set hyperparameters that prove be better suited task, comparing quality seasonal autoregressive integrated moving average models. Furthermore, an innovative representation seasonality given by means embedding layer produces mapping latent space, parameters such being tuned using predictions. Finally, present some managerial implications, putting into evidence research limitations future opportunities.