The radial vibration performance of the feeding system with nut screw drive can affect directly the accuracy and reliability of machine tools, and the radial vibration performance was primarily influenced by the radial stiffness and damping of bearing, nut and guideway. Currently, there is no a mature algorithm in the identification of the parameters at each interface of parts. Therefore, the transfer matrix method based on Timoshenko beam were used to predict frequency response functions of the feeding system, and improved Particle Swarm Optimization (PSO) algorithm was used to determine the radial stiffness and damping of bearing, nut and guideway by minimizing theory and experimental results. Comparison results show that the predicted and experimental results show a good agreement, which demonstrate the proposed identification method is well valid in the recognition of the interface parameters at the feeding system.