Short Term Public Transit Dispatch Model Using State Space Neural Networks

This paper discusses the solution of proposed bus dispatching in short terms using state space neural networks (SSNN). Instead of treating the neural network as a “black-box” model, the design of SSNN can explicitly reflect the relationship that exists in physical public transit system. It allows the interpretation of neuron weights and structure in terms of the inherent mechanism of the network process with clear physical meaning. Model performance is tested by a densely used public transit data in Nanjing. BP neural networks and ARMA model are investigated to compare the performance of the model. Results of the comparisons indicate that in short terms SSNN model can adjust bus departing interval based on passenger flow space and time variations and predict bus dispatching with satisfying effectiveness, robustness and reliability.