Magnetic shape memory alloy actuator have hysteresis characteristics, in order to describe the hysteresis nonlinearity of magnetic shape memory alloy actuator and study the methods of eliminating the hysteresis nonlinearity of magnetic shape memory alloy actuator, the Krasnosel'skii-Pokrovskii (KP) model is firstly established in this paper. Secondly, BP neural network and adaptive linear neural network are applied to identify the density functions of the KP model respectively. Finally, it is obtained that the modeling accuracy of the KP model is 0.37% by BP algorithm and the modeling accuracy of the KP model based on adaptive linear neural network is 0.19% through the simulation results, which prove that the KP model can characterize the hysteresis nonlinearity of magnetic shape memory alloy actuator.