Science And Technology Application Of BP Neural Network In Short-term Forecast Of Passenger Flow Of Rail Transit

Urban rail transit with its large capacity,fast,punctual,safe,environmentally friendly and become the most effective means to solve the traffic congestion.The short-term forecast of rail transit passenger traffic is the basis for the timely adjustment of the operation plan of the passenger transport department,and it is also an important basis for evaluating the level of rail transit service.Based on the research of BP neural network,this paper establishes a short-term forecasting model of rail transit passenger traffic based on BP neural network,and uses the LM algorithm to improve the BP neural network.Using the Yushan Road Station of Nanjing Metro Line 10,Road Station,Longhua Road Station and Riverside Road Station.The results show that the prediction results of LM-BP neural network are significantly lower than those of the original BP neural network,and the prediction accuracy is significantly improved by the average percentage error(MAPE)and mean absolute deviation(MAD).In order to overcome the defects that LM-BP neural network is easy to fall into local minimum,this paper uses GA(Genetic algorithm)and particle swarm optimization(PSO)to optimize LM-BP algorithm respectively.Based on this,And PSO-LM-BP Neural Network for Short-term Forecasting Model of Rail Transit.Genetic algorithm can optimize the initial weights and thresholds of LM-BP neural networks,and obtain the optimal solution and then replace them into the original neural network.The particle swarm algorithm is the initial weight and threshold of LM-BP Randomly initialize the particles,the iterative way to constantly update their own speed and location,and ultimately find the optimal solution and the optimal position,and then the weight and threshold into the original neural network to obtain the optimal solution.The results show that the LM-BP neural network based on the genetic algorithm and the particle swarm optimization algorithm can be used to simulate the MAPE and MAD in the MAPE and MAD The prediction results of the two LM-BP methods are reduced and the prediction accuracy is further improved.