Research On Improving BP Neural Network In Forecasting Airline Sales Volume

With the rapid development of the civil aviation industry,travel by air has become the first choice for many residents of China,and the number of ticket sales points and the sales volume of air tickets are increasing.Accurate forecasting of ticket sales is more conducive to the planning of the ticket sales market.How to improve the forecast accuracy of ticket sales has important social significance.Domestic and foreign scholars have made in-depth research on the sales volume forecast in the sales field.On the basis of previous studies,finding more stable and accurate prediction algorithms has become the focus of many experts in recent years.BP neural network is a widely used data mining prediction algorithm.The initial weights and thresholds of the traditional BP neural network are randomly generated,and then inversely adjusted by the gradient descent method until the end of training.If these two initial parameters are not properly selected,it is easy to cause local optimization of the network and slow convergence.In order to effectively improve the prediction performance of BP neural network,this paper proposes a hybrid optimization algorithm(GA_HS)of adaptive harmony algorithm(HS)and genetic algorithm(GA)to optimize the initial weight and threshold parameters of BP neural network..The prediction model is compared with the prediction accuracy of the improved and improved BP neural network to prove the feasibility of the improved algorithm.The experimental results show that the GA_HS_BP model based on GA_HS algorithm optimizes the prediction accuracy of GA_HS_BP model,but the fitness function value of GA_HS algorithm decreases rapidly at 90 times,and its convergence speed is slow.In order to further improve the convergence speed of the improved algorithm and better optimize the BP neural network,the hybrid weighting algorithm(BFA_FPA)of the bacterial foraging algorithm(BFA)and the adaptive flower pollination algorithm(FPA)is proposed to give the initial weight of the BP neural network.And the threshold is optimized,the BFA_FPA_BP model is established,and the prediction contrast experiment is completed.The final experimental results show that the BP neural network optimized by BFA_FPA algorithm has further improved the prediction accuracy and search speed,and proves the effectiveness of the improved algorithm.Finally,the BFA_FPA_BP model with relatively good predictive performance is used to predict the ticket sales volume.