Forecast Of Short-term Bus Passenger Flow Based On IC Card Data

With the accelerating process of urbanization in China,the urban traffic congestion has become more and more serious.In order to solve this problem,in December 2012,the State Council announced “The priority development of public transport in urban guidance”.The national community has further established the priority development strategy of urban public transport system.In the implementation of public transport priority strategy,public transport scheduling is the core of the bus operation and management,which can effectively ease the contradiction between capacity and demand,and balance the interest of passengers and bus enterprise.Due to the widespread use of bus IC card,it provides a new bus passenger data collection method,which can be used for real-time analysis and short-term forecast.This paper uses IC Card data to analyze the bus passenger flow distributions and characteristics,and then through the use of intelligent algorithms for short-term bus passenger forecast.The main contents of this paper are as follows:(1)Introduce the bus IC card data,and demonstrate the feasibility and application of bus IC card data analysis.Using the quartile method to judge whether there is any abnormality in the bus data.Analyze the characteristics of short-term passenger flow,and provide the basis for short-term forecasting.(2)Based on the distribution and characteristics of bus passenger flow at different times,the passenger flow data of adjacent times,of adjacent days and of the adjacent weeks are determined as the model input.The BP neural network algorithm and particle swarm optimization algorithm are used to build the forecast model.In the establishment of BP neural network prediction model,the different input layer,the transfer function and the number of hidden neurons are compared and analyzed to determine the ideal combination and improve the short-term prediction accuracy.Particle swarm optimization algorithm are further optimized to BP neural network model.The results of the two predictive models are evaluated by using the mean square error and the average absolute percentage error.The results show that the two models meet the short-term prediction requirements,and the particle swarm optimization algorithm can improve the prediction accuracy.