How Much Longer? Estimating Bus Arrival Times with Predictive Models

Abstract

Predicting bus arrival times accurately is essential for improving urban mobility and enhancing public transportation services. Delays and uncertainty in bus schedules can lead to passenger frustration and inefficient travel planning. In this context, providing real-time, reliable arrival time estimates can help commuters reduce waiting times and make informed decisions. This work explores different predictive approaches, including Linear Regression, ARIMA, Long Short-Term Memory (LSTM), and gated recurrent units (GRU), to estimate bus arrival times based on real-world bus GPS data from the city of Tandil (Buenos Aires, Argentina). Experimental results demonstrate that deep learning models, particularly LSTM, can significantly outperform traditional approaches, highlighting their potential to optimize public transportation systems. In addition to developing predictive models, we provide a mobile application that integrates the prediction models, offering users real-time information on bus arrival times.