Application of LDHA-BP in Prediction of Atmospheric PM2.5 Concentration

LDHA-BP algorithm is designed for lack of using BP neural networks to predict atmospheric concentrations of PM2.5. In this algorithm, the process of solving the optimal weights and thresholds of BP neural networks will be transformed into the process of finding the optimal location of the dolphin species predator's, by introducing Leader Dolphins Herd Algorithm (LDHA).The algorithm effectively combines the good generalization ability of BP neural network neural and the global optimization ability, local search capability of LDHA. Test data for PM2.5 from a monitoring point in Wuhan to predict the PM2.5 concentrations in future. The experimental results show that LDHA-BP is faster, more accurate to predict the PM2.5 concentrations and it has good practical value.