High-frequency ground wave radar is an important means of detecting moving targets on the sea, among which sea clutter is the primary factor that influences the function of sea surface target detection. For the sake of enhancing the prediction precision of sea clutter and validly restrain the sea clutter, this paper proposes a sea clutter prediction model (MGPALO-RBF) based on the improved ant lion algorithm to optimize the RBF neural network. In order to ameliorate the weak point of the ant lion algorithm that it sink easily into the local optimum and the convergence velocity is slow, a disturbance factor is added to increase the activity and variety of the population during the process of stochastic walk of the ants; multiple elite dynamic guidance mechanisms are put forward to intensify the searching capability of the algorithm in the prophase and the exploit ability in the later stage; Gaussian difference mutation is put forward on the poor ant lion in the population to enhance the convergence velocity of the algorithm. The simulation results make clear that the mended ant lion algorithm has higher convergence precision and constriction speed in the contrast algorithm, and the optimized model has a better effect in sea clutter prediction.