Stiffness variation is crucial to the performance of suspension systems. However, the existing suspension systems can only vary the stiffness by means of indirect ways and then leads to the difficulty in achieving optimal stiffness. To cope with this problem, a multimodal quasi-zero-stiffness (QZS) mechanism is proposed to ensure the optimal stiffness of suspension system by varying the dynamic stiffness (DS). Besides, a data-driven DS variation algorithm is developed to address the modal transition of the QZS mechanism with unknown knowledge. Moreover, the decoupling performance of the modal transition is guaranteed by designing a constraint-based decoupling scheme. By combining the decoupling scheme, a data-driven strategy is proposed to approximate the optimal solution of the DS variation algorithm. Moreover, the superiority of the proposed QZS mechanism is verified through the comparison with the general active suspension system.