Based on support vector machine theory, a generalized predictive controller is designed for turboshaft engine. Using dynamic response data, through support vector machine regression, the turboshaft engine nonlinear model is established. By linearizing the nonlinear model at each sampling cycle, the prediction model of turboshaft engine is constructed. According to generalized predictive control algorithm, the optimal control signals are obtained. Simulation results show that the control system can follow the change of collective pitch very well, the speed of power turbine can back to given value fast and possess small overshoot.