Predictive model of maritime port transport of cargo mobilized incabotage and transshipment

Abstract

Although there have been some advances in maritime transport and port operations, it remains necessary to understand the transport sector related to cargo movement through the various authorized ports in Colombia, as this constitutes a strategic factor for international trade. Despite the fact that monitoring is carried out through reports from public and private port companies via the Transport Supervision Information System (VIGIA), significant gaps persist both in the understanding of coastal maritime transport (cabotage) and in the recording of cargo tonnage moved by port zone, port traffic, type of traffic, type of cargo, and port company. These aspects are essential for strategic decision making. The purpose of this study was to analyze the types of port cargo moved in cabotage and transshipment operations Colombian ports. To this end, the CART (Classification and Regression Trees) methodology—an automatic supervised learning algorithm—was applied. The secondary data used came from datasets provided by public and private port companies, totaling 7,139 records corresponding to public and private port zones in the country between 2018 and March 2024. The results present a predictive model based on machine learning for cargo type classification, using a decision tree with a categorical target variable that distinguishes between bulk coal cargo, containers, general cargo, liquid bulk, and solid bulk other than coal. The predictive variables (features) considered were cabotage and transshipment. It is concluded that the decision tree model, in its current configuration, shows limited performance in classifying cargo type when using only these features.
Unlike other studies conducted in Colombia or the region—which focus mainly on logistics optimization, terminal management, or port efficiency analysis through quantitative indicators— this research incorporates an artificial intelligence and machine learning approach. The use of the CART methodology represents a methodological innovation, as it enables the identification of port cargo classification patterns based on categorical and supervised variables, offering a more predictive rather than descriptive perspective on the behavior of national maritime traffic.