The car door plays a critical role in the overall structure of a vehicle, requiring attention to achieve optimal mode and stiffness performance in design. Currently, the analysis of car door modes and stiffness heavily relies on finite element analysis methods. A significant challenge lies in the high cost and lengthy cycle involved in the analysis and optimization process. This study focuses on the prediction of the structural performance of automotive doors based on the back propagation (BP) neural network. A simulation analysis model was developed for a car door, followed by the application of an optimal Latin hypercube design to ensure evenly distributed data samples. A BP neural network model was constructed and trained to accurately predict the door’s mode and stiffness, achieving an error rate of less than 5%. The combination of the BP neural network and optimal Latin hypercube design provides a reliable and efficient method for the prediction of automotive door structural performance, offering valuable guidance for the design and improvement of car door structures in the automotive industry.