A Method for the Dynamic Distribution of Special Population Prediction Based on GA-BP

Abstract

In order to predict the dynamic distribution of special population effectively, it is very necessary to establish an early warning population system. In this paper, a Back Propagation (BP) neural network quantitative prediction optimization model based Genetic Algorithm (GA-BP model) is proposed. Firstly, the initial weight and the thresholds of the BP network are optimized by GA. Then, the BP network starts training the related data of special population, which will be used to predict the distribution of special population. Finally, some standard errors are used to verify the proposed GA-BP model, and the simulation results demonstrate that GA-BP model outperforms the GA both in prediction accuracy and convergence. According to the average value of the MSE index numerical results of the defined function, the following numerical results are obtained: for the model established, the average value of the GA-BP function is 10.3273 compared with the average of 20.8815 of the BP model, and has better performance and more stable performance. In addition, the improved model can be adopted to show the early warning of the population distribution, which has a certain reference value for urban traffic managers.