Bus-arrival time prediction using bus network data model and time periods

Abstract

Real-time information system about bus arrival and departure times at bus stops is one of the key elements of high-quality public transport. To provide accurate prediction of bus arrival at the bus stop the information of current position of the bus and bus travel time to the target stop are required. This paper investigates the bus velocity during different time periods and the impact of the bus network data model on the real-time prediction of arrival times at bus stops. The proposed model classifies bus travel times into time periods with respect to the historical data and the data model of the bus network. Discussion of four types of data model is presented, such as: a data model defined by bus stops and junctions of the roads, a data model defined only by bus stops, a data model which addresses the individual parts of the network in relation to the potential barriers that affect the travel speed of buses, and a data model with fixed-length links of the bus network. The analysis is performed in two different environments, the city of Ljubljana and the city of Maribor. The travel times are classified according to the average travel speed into four time periods: morning, afternoon, weekend and off-peak periods. Simulations showed that both the data model of the bus network and classifying runs into time periods affect the accuracy of predictions of bus arrival times at bus stops.