Transit OD Inference from Mobile Ticketing Data

This research develops machine learning frameworks to infer complete origin-destination-transfer matrices from boarding-only mobile ticketing data in regional bus networks with 40+ routes.

Research Overview

The project addresses the critical challenge of understanding complete passenger travel patterns when transit agencies only collect boarding data. Many regional transit systems in the US use mobile ticketing platforms that record when and where passengers board, but not where they alight. This research creates cost-effective methods to infer full journey patterns without expensive smart card infrastructure.

Methodology

Passenger Typology Clustering

• Developed clustering framework using hierarchical methods and dynamic time warping (DTW)
• Identified distinct behavioral patterns across 40,000+ weekly riders
• Captured spatiotemporal heterogeneity in passenger travel patterns

Trip Chaining with Spatial Error Correction

• Built trip chaining model incorporating passenger typology and seasonal variations
• Integrated gradient boosting machine (GBM) for spatial error correction
• Reduced Mean Absolute Error (MAE) by 70% and Symmetric Mean Absolute Percentage Error (SMAPE) by 85%

Origin-Destination-Transfer (ODX) Model

• Advanced model inferring complete trip patterns including transfers
• Leverages iterative proportional fitting (IPF) and machine learning
• Scales partial AFC-based OD estimates to full transit network population

Impact and Applications

• Scalable methodology for transit agencies without smart card infrastructure
• Directly applicable to US agencies using similar mobile ticketing platforms
• Validated with comprehensive survey data from Pioneer Valley Transit Authority
• Enables data-driven network planning and service optimization

Funding

Pioneer Valley Transit Authority (PVTA), 2021-2023 (PI: J. Oke)

Key Publications

Origin-destination inference in public transportation systems: A comprehensive review (2023)

• Published in International Journal of Transportation Science and Technology
• Comprehensive review of state-of-the-art OD estimation methods
• Analysis of transition from survey-based to big data-driven techniques

Extracting Spatiotemporal Bus Passenger Trip Typologies from Noisy Mobile Ticketing Boarding Data (2023)

• Published in Data Science for Transportation
• Novel clustering framework using hierarchical clustering and DTW
• Identification of distinct passenger behavior patterns

Enhanced Seasonal Typology-Informed Transit Trip Chaining via Mobile Boarding and Survey Data (2024)

• Published in Data Science for Transportation
• Trip chaining model with gradient boosting for spatial error correction
• Achieved 70% reduction in MAE and 85% reduction in SMAPE

A Novel Origin-Destination-Transfer Model (Under Review)

• Integration of trip chaining, typology, and spatial error correction
• Complete inference of passenger journeys including transfers