Improving bus arrival time predictors using only public transport API data

ABSTRACT

Accurate prediction of bus arrival times can greatly benefit public transport users, allowing them to better plan their journeys in cities. The usual Expected Time of Arrival (ETA) estimators provided to citizens use all the information available to the bus service provider (vehicle position, traffic, etc.); in this paper we propose a procedure to improve these estimators that relies solely on historical ETA records provided by public transport councils through application programming interfaces (APIs). This improvement is achieved by means of a machine learning scheme that predicts and corrects the systematic errors of the available ETA estimators. Significant improvements in terms of error mean and standard deviation are achieved for the Madrid and Paris bus fleets. These robust results and the fact that the proposed scheme uses only historical and online information provided by APIs, without requiring the cooperation of the service provider, support the suitability of the proposed method for general public benefit applications toward the sustainability of cities.

KEYWORDS:

• Public transport
• vehicles
• intelligent transportation systems
• arrival time estimation
• machine learning

Acknowledgments

The authors are grateful for the help and support of the researchers and those responsible for the Cátedra Cabify (Cabify Chair) at ETSIT-UPM.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Research mainly supported by Cabify Matriz, S.L. through the Cátedra Cabify (Cabify Chair) at ETSIT-UPM, and by the Ministerio de Ciencia e Innovación of Spain grant PID2020-112502RB/AEI/10.13039/501100011033; this work is also a part of the project TED2021-129189B-C21/TED2021-129189B-C22, financed by MCIN/AEI/10.13039/501100011033 and the European Union “NextGenerationEU”/PRTR.