Doing shipping well with predictions: machine learning-based port congestion analysis

Container ports are crucial hubs for global trade and are essential for the smooth functioning of the supply chain. Nevertheless, the congestion currently observed in the world’s major container ports could potentially result in a catastrophic global shipping crisis. Despite this, few studies have used machine learning to predict container congestion accurately. Therefore, this study collects daily shipping data from 1 January 2016, to 24 March 2023, totaling 21120 observations. In light of this, we developed a deep learning integrated model (CNN-BILSTM-AM) that combines an attention mechanism (AM), a convolutional neural network (CNN), and a bi-directional long-term and short-term memory network (BILSTM) to predict container port congestion. The outcomes demonstrate that the integrated CNN-BILSTM-AM model effectively captures port congestion’s nonlinear and time-varying characteristics. With excellent adaptability to random sample selection, data frequency, and sample structure breaks, its prediction accuracy (with a value of 95.57%) significantly outperforms that of the traditional and single models. This research contributes to the advancement of the application of machine learning in the prediction of port congestion risk and the issuance of ocean disaster warnings. The model provides decision support and risk management tools for port managers, logistics companies, and government policymakers.