Travel pattern-based bus trip origin-destination estimation using smart card data

Area of Science:

• Transportation Science
• Data Science
• Urban Planning

Background:

• Smart card data is vital for public transit origin-destination (O-D) matrix generation.
• Current methods struggle with unlinked trips lacking destination information.
• Existing trip chain methods have limitations in estimating all destinations.

Purpose of the Study:

• To develop a novel method for estimating unlinked trip destinations using smart card data.
• To leverage temporal travel patterns and historical passenger data for improved accuracy.
• To address limitations of current trip chain methods in public transit analysis.

Main Methods:

• Utilized k-means clustering to group passengers based on travel behavior.
• Employed Gaussian mixture models to estimate time-of-day travel patterns per cluster.
• Formulated travel patterns to predict passenger destinations from smart card data.

Main Results:

• The proposed method successfully estimated destinations for previously unlinked trips.
• Destination matching improved from 60.0% (existing methods) to 74.9%.
• An additional 37.2% of unlinked trips had their destinations matched.

Conclusions:

• The developed method significantly enhances the accuracy of public transit O-D matrices.
• This approach provides a robust solution for handling unlinked trips in smart card data.
• Improved O-D matrix accuracy supports better transportation planning and operations.