To overcome the tracking control problem for a class of strict-feedback nonlinear system with model uncertain, an adaptive fuzzy backstepping predication control algorithm which can make the closed-loop system stable and minimize the receding horizon guaranteed cost on-line is proposed. Fuzzy logic systems are employed to approximate the unknown term in the design process. As the adaptive parameter are directly used to estimate the norm of the optimal approximation weight vector, only one parameter need to be tuned on-line. Considering the property of the fuzzy basis function, the designed control laws and adaptive laws do not contain the fuzzy basis function term. Theoretically, it is proved that the using the constructed controller can guarantee that all signals in closed-loop are semi-globally uniformly ultimately bounded, and the tracking error convergence to a small neighborhood of the origin. As the form of the controller designed in this way is simplicity and the computation is small, this control strategy is easily realized in practice. Finally, the simulation results demonstrate the feasibility of the proposed scheme.