A Comparative Study of Traditional Excel Forecasting Methods and Machine Learning Techniques for Freight Volume Prediction in Uzbekistan

Accurate forecasting of freight volumes is essential for effective transportation planning and infrastructure development. Previous research on Uzbekistan’s railway sector primarily relied on single-method approaches, either using traditional statistical tools or machine learning techniques. This study adopts an innovative dual-method framework, combining Excel-based models—such as regression equation, exponential smoothing, and moving average—with advanced machine learning techniques, including decision tree, random forest, gradient boosting, and extreme gradient boosting. Freight shipment data and socio-economic variables, such as gros domestic product and operational railway length. Model performance was evaluated using root mean square error and mean absolute percentage error. The regression equation model demonstrated exceptional precision with a mean absolute percentage error of 0.001%, though its simplicity raised concerns about overfitting and limited scalability. Meanwhile, machine learning models showcased superior robustness and generalization capabilities, achieving low and balanced error rates, making them more suitable for capturing complex, non-linear relationships in freight dynamics. According to the compound annual growth rate projection, freight volumes are expected to increase significantly, reaching 106 million tons by 2030. This underscores the growing importance of strategic infrastructure investment, modernization, and policy interventions to accommodate future demand. The findings provide valuable insights for policymakers and transportation planners, offering a practical and comprehensive framework for sustainable development in Uzbekistan’s railway sector. This study aims to lay a foundation for informed decision-making and long-term growth planning by leveraging a mix of traditional and modern forecasting approaches.