Comprehensive Cost Calculation in Road Freight Transport Using Artificial Intelligence and Machine Learning: Methodology and Applications in Logistics. Case of Colombia Freight.

Abstract

Accurate prediction of freight transport costs is critical for logistics planning, infrastructure policy, and supply chain optimization. Conventional cost estimation approaches rely on deterministic or linear econometric models that often fail to capture nonlinear interactions among macroeconomic indices, operational logistics variables, and route-specific characteristics. This study proposes a machine learning framework for predicting freight transport costs across origin–destination routes and cost components using historical cost indices and operational transport records.

The model integrates temporal cost dynamics, spatial logistics attributes, and cost category decomposition within a supervised nonlinear regression architecture. Empirical evaluation demonstrates that ensemble-based learning models significantly outperform linear benchmarks in predictive accuracy and robustness. The proposed framework enables route-level cost forecasting, interpretable cost decomposition, and uncertainty quantification, providing decision-support capabilities for logistics operators and policy makers.

Road freight transport presents technical, economic and regulatory challenges that require advanced methodologies to optimize costing. The incorporation of artificial intelligence and machine learning allows to dynamically model and adjust fixed, variable, additional components and monetized externalities, using cleaned historical data, real-time monitoring with IoT sensors and GIS systems, along with predictive maintenance management. This approach integrates contextual variables such as weather conditions, road conditions, abrupt variations in demand and urban restrictions, improving operational efficiency and reducing socio-environmental impacts.

Practical results show reductions in variable costs and penalties, increases in effective vehicle occupancy and reduction of polluting emissions. The methodology facilitates adaptive decision-making in fragmented and urban logistics chains, promoting sustainable and competitive management of land freight transport.