Travel Time Prediction Between Bus Stops Based on an Optimized BP Neural Network

With the acceleration of urbanization, the instability of bus journey times has a negative impact on the quality and attractiveness of bus services. In order to improve the prediction accuracy of bus inter-station travel time, this study proposes an optimized BP neural network-based bus inter-station travel time prediction model. Six key factors, such as distance between stops, time period, and number of signals, are selected as input variables in the study, and Particle Swarm Algorithm (PSO) and Genetic Algorithm (GA) are introduced to optimize the initial thresholds and weights of the BP neural network. The experimental data come from some sections of 20 bus routes in Nanjing, and the bus operation state is simulated by VISSIM software, 680 sets of sample data are collected, which are randomly divided into 540 training sets and 140 test sets. The experimental results show that compared with the traditional BP neural network model, the optimized model has a significant improvement in training efficiency and prediction accuracy, with the number of its training iterations reduced from 22 to 14, the mean absolute percentage error (MAPE) of the prediction model reduced from 5.1847% to 2.1433%, and the mean square error (MSE) reduced from 36.846 to 20.977. The model is useful for the prediction of travel time between bus stops and operation management, provides an effective reference.