BusTr: Predicting Bus Travel Times from Real-Time Traffic

TL;DR

BusTr is a machine learning model that predicts bus delays from traffic forecasts, significantly outperforming previous models and generalizing well across diverse transit systems.

Contribution

The paper introduces BusTr, a neural sequence model that improves bus delay prediction accuracy and stability over existing methods, with enhanced generalization capabilities.

Findings

30% reduction in MAPE compared to DeepTTE

Improved training stability and performance

Strong generalization across different transit systems

Abstract

We present BusTr, a machine-learned model for translating road traffic forecasts into predictions of bus delays, used by Google Maps to serve the majority of the world's public transit systems where no official real-time bus tracking is provided. We demonstrate that our neural sequence model improves over DeepTTE, the state-of-the-art baseline, both in performance (-30% MAPE) and training stability. We also demonstrate significant generalization gains over simpler models, evaluated on longitudinal data to cope with a constantly evolving world.