The Panama Canal is one of the most important logistics infrastructures in international maritime trade, whose efficiency has a direct impact on the performance of global supply chains. The objective of this study is to analyze the behavior of transit time and waiting time of ships, as well as to evaluate their predictive capacity based on technical, operational, meteorological and hydrological variables. The research was developed under a quantitative approach using historical data corresponding to the period 2019–2023, which were subjected to processes of cleaning, integration, temporal alignment and exploratory analysis. Subsequently, machine learning models, including Random Forest, XGBoost, KNN, LightGBM and neural networks, were implemented using temporal validation. The results show that the transit time presents a stable behavior and a concentrated distribution, mainly influenced by the size and technical characteristics of the vessel. In contrast, the waiting time shows greater variability, a skewed distribution and an increasing trend, even in a context of reduced number of transits. Likewise, the models perform better in the prediction of transit time than in that of waiting time, which suggests the influence of unobserved operational factors. It is concluded that the congestion of the Canal does not respond only to the volume of traffic, but also to the interaction between demand, effective operational capacity and external restrictions of the system.

Keywords

Panama Canal, transit time, waiting time, neural networks, operating times