An Adaptive Algorithm for Public Transport Arrival Time Prediction Based on Hierarhical Regression

In this paper we consider a problem of public transport arrival time prediction for a large city in real time. We propose a new prediction algorithm based on a model of an adaptive combination of elementary prediction algorithms, each of which is characterized by a small number of adjustable parameters. Adaptability means that parameters of the constructed combination depend on a number of control parameters of the model, which includes the following factors: weather conditions, traffic density, driving dynamics, prediction horizon, and others. Adaptability is achieved by the use of a hierarchical regression (similar to a regression tree). The proposed arrival prediction algorithm has been tested with the data of all the public transport routes in Samara, Russia.