Abstract: To thoroughly investigate the dynamic process of modeling the relationship between bearing load and displacement driven by transfer learning, a specific multi-supported shafting was selected as the research object. The key elements of modeling were systematically analyzed, and an orthogonal experimental scheme was designed for three critical factors: the accuracy of transferred knowledge, the adaptability of transfer strategies, and the precision of target data samples. Modeling research based on transfer learning was conducted accordingly. The results indicate that the ranking of factor importance is as follows: accuracy of transferred knowledge > adaptability of transfer strategy > precision of target data samples. During modeling, priority should be given to the accuracy of transferred knowledge and the adaptability of transfer strategies to rapidly align with the target domain. Additionally, maintaining the error range of target data samples within 4% ensures the relative stability of the modeling process. This study provides theoretical support for constructing the relationship model between bearing load and displacement in engineering applications, offering significant value in reducing testing costs and enhancing modeling efficiency.