To analyze the rollover safety, finite element models were established for the electric bus body frame, rollover simulation platform, living space, and bus rollover. The strength and stiffness of the body frame were calculated under four typical working conditions considering the main low-order elastic modal characteristics. The results indicate that the initial body frame of the electric bus satisfies the required structural strength, stiffness, modes, and rollover safety, and it has great potential for lightweight design. Sensitivity and structural contribution analyses were performed to determine the design variables for lightweight optimization of the body frame, and a mathematical model was established for multi-objective collaborative optimization design of the electric bus. Then, the radial basis function neural network was used to approximate the optimization model. Besides, the accuracy of the approximate model was verified, and the non-dominated sorting genetic algorithm II was employed to determine solutions for the lightweight optimization. Compared with the initial model, the mass of the optimized model is reduced by 240 kg (9.0%) without any changes in the materials of the body frame.