Optimization of BP artificial neural network regression prediction model based on Improved Sparrow search algorithm with Sine chaotic mapping and its application

Sparrow search algorithm is a new meta-heuristic algorithm developed in recent years, which mainly simulates the foraging behavior of sparrows, and has higher local search ability and convergence performance. However, when the algorithm approaches to the global optimum, with the decrease of population diversity, it is facile to be included in the local optimum. After comparing the four algorithms, it is concluded that SSA has advantages over PSO in search accuracy and convergence speed. The sparrow population was initialized by sinusoidal chaotic map, and the individuals with higher fitness ranking were selected as the initial population, and the initial weight and threshold in BP artificial neural network were optimized. Finally, the mean absolute percentage error (MAPE) of Sine-SSA-BP model is 17.5% compared with the traditional BP artificial neural network, which is greatly optimized compared with the standard BP neural network of 33.7%