According to the strong coupling between the electromagnetic torque and radial suspension force, and the difficulty of establishing a precise mathematical model of bearingless induction motor, least squares support vector machine is proposed in this paper. The inverse model approximated by the LS-SVM is connected with the original system, decoupling a complex nonlinear multivariable system into four relatively independent pseudo-linear sub-systems of single input and single output. At the same time, fuzzy PID control strategy is introduced to the pseudo linear system to ensure the robustness and anti-jamming ability of the control system. Then the performance is simulated with Matlab/ Simulink. The theory research and simulation experiment have validated that the decoupling control of inverse system based on least squares support vector machine method can be successfully realized, which between the electromagnetic torque and radial suspension force, and the good performance of dynamic and static state of system can be also obtained.