Modeling Shippers' Behavior: Hybrid Statistical Method and Monte Carlo Simulation for Imbalance Data

Abstract: Freight transportation is the backbone of the U.S. economy, seamlessly moving vast quantities of goods, from raw materials to consumer products, across a complex multimodal network. This intricate system not only powers economic growth but also ensures that supply chains remain efficient and responsive to dynamic market demands. This study focuses on calibrating shippers' utility functions, which model the decision-making processes for selecting transportation modes and routes. To do this, we use data from the Commodity Flow Survey (CFS), a key resource that provides insights into freight movement across the U.S. However, CFS data reveals a significant imbalance in transportation modes, with some modes being underrepresented. This imbalance poses a challenge to traditional modeling techniques, which can lead to skewed or inaccurate predictions. We propose a hybrid approach that integrates machine learning and Monte Carlo simulation to address the issue of imbalanced data. The Monte Carlo simulation demonstrates how this imbalance affects prediction accuracy, particularly in the context of freight mode choice. By employing class weighting techniques within machine learning models, we correct for the underrepresentation of certain modes, improving the accuracy and reliability of the predictions. This enables better forecasting of how shippers re-optimize their transportation decisions when faced with disruptions, such as port closures or capacity changes. The results of this research provide critical insights for transportation planners, helping them design systems that are more resilient to disruptions. By improving predictive accuracy through advanced modeling techniques, this study offers a roadmap for enhancing the reliability and resilience of multimodal transportation systems in the face of ongoing global challenges. This work underscores the importance of addressing imbalanced data in freight transportation modeling, offering practical solutions for better decision-making in the sector.