In order to efficiently and accurately predict freight forwarders' transportation modes preferences between China and Europe during major emergencies, as well as to uncover the relevant factors influencing freight forwarders' choices, the stated preference method was employed to survey freight forwarders. Additionally, considering the influences of transportation and cargo attributes, decision trees, logistic regressions, and random forest prediction models were constructed to forecast the selection behavior of freight forwarders. The prediction results of the machine learning model and the discrete choice model were comprehensively compared through four evaluation metrics: accuracy, precision, recall, and F₁ score. Furthermore, the random forest algorithm was utilized to rank the importance of attributes influencing freight forwarders' transportation mode choices during different stages of the pandemic. The study results demonstrate that the prediction accuracy of all three machine learning models is higher than that of the discrete choice model. Among them, the random forest model exhibits superior prediction accuracy compared to the decision tree and logistic regression models in addressing the choice of Sino-Europe container transport modes, making it more suitable for this problem. Regarding influencing factors, during stable periods, cargo attributes are identified as the most important factors. When major emergencies occur, freight forwarders place greater emphasis on the threshold delay time. Furthermore, the destination and value of the cargo are found to have significant impacts on the choice of Sino-Europe container transport modes. The study proposes an accurate analysis of the decision-making mechanisms guiding freight forwarders' mode choice behavior during major global emergencies. Furthermore, it is utilized by shipping companies and operators of the China Railway Express to gain a deeper understanding of the preferences and decision-making factors influencing freight forwarders. The insights derived from this study are considered a solid basis for effectively responding to similar emergency situations.