Research on Vehicle Retention Rate Prediction Combined with Pre-Trained Language Model

The chapter delves into the complexities of predicting used vehicle retention rates, traditionally handled by statistical models and domain expertise. It introduces the application of pre-trained language models, such as BERT and ELECTRA, which leverage large-scale language data to enhance prediction accuracy and generalization. The research compares these advanced models with traditional methods, demonstrating their superior performance in handling complex datasets and variable market environments. The study also includes a detailed dataset construction process, normalization techniques, and experimental results, providing a thorough analysis of the influencing factors and predictive capabilities of these models. The chapter concludes with a comparative discussion of the three models, highlighting the strengths and potential of pre-trained language models in the automotive industry.