Abstract: The modified observation noise in Particle Filter (PF) algorithm was proposed to solve the problem of difficulty in obtaining observation noise values in PF algorithm. The Compressed Monte Carlo (CMC) algorithm was introduced. The CMC algorithm and the Particle Swarm Optimization Particle Filter (PSO-PF) algorithm were combined to obtain the improved Particle Swarm Optimization Particle Filter (CMC-PSO-PF) algorithm based on the CMC algorithm, so as to improve the computational efficiency of the PSO-PF algorithm. The application of the CMC-PSO-PF algorithm in the dynamic response prediction of the prefabricated subway elevated station structures was carried out. The research results indicate that compared with the observation noise variance R set at a constant value, the modified observation noise has a clear value equation, good accuracy, and stronger applicability; that the computational accuracy and efficiency of the PSO-PF algorithm are better than those of the PF algorithm; that by setting reasonable parameters, the CMC-PSO-PF algorithm can achieve the computational accuracy of the PSO-PF algorithm and reduce the calculation time by 19.9% to 43.2%; that as the number of framework layers increases or the measurement noise increases, the computational accuracy of the PSO-PF algorithm and the CMC-PSO-PF algorithm gradually decreases; that the CMC-PSO-PF algorithm has good computational accuracy in the structural stiffness parameter identification of the prefabricated subway elevated station structures; and that the shear dynamics model of the prefabricated subway elevated station structure established upon the parameter identification results of the CMC-PSO-PF algorithm can predict the structural dynamic response relatively well.