Abstract: The study of shipper behavior is essential for understanding mode choices within the context of national commodity flow analysis. Many shipment characteristics are important for behavioral analysis, yet obtaining relevant data remains a challenge for modeling. This thesis develops an interpretable mode choice model using the 2017 Commodity Flow Survey (CFS) Public Use File (PUF) data. A multinomial logit model is calibrated via maximum likelihood estimation. This structure represents shipment characteristics through indicators that influence the decision maker's utility.

The model is assessed across five major US freight corridors using standard performance measures, viz., confusion matrices, disaggregate accuracy, and out-of-sample errors. Evaluated model achieved about 80% disaggregate accuracy and a 1.0% aggregate RMSE while maintaining generality. Sensitivity checks confirm the stability of these parameters.

The proposed discrete-choice framework yields precise predictions and meaningful insights that reinforce applications in freight network analysis, demand planning, and policy interventions. This study provides a strong baseline for higher-resolution network data and advanced choice structures.