Bicycle Data-Driven Application Framework: A Dutch Case Study on Machine Learning-Based Bicycle Delay Estimation at...

Area of Science:

• Transportation Engineering
• Data Science
• Urban Planning

Background:

• Nations are increasingly adopting data-driven methods for transportation system evaluation.
• The Netherlands has established protocols for bicycle traffic counting and GPS data collection.
• Accurate estimation of bicycle delays at signalized intersections is vital for performance assessment.

Purpose of the Study:

• To develop a generic framework for analyzing cycling data.
• To estimate average bicycle delays at signalized intersections using machine learning.
• To assess the feasibility of using sparse GPS data for delay estimation.

Main Methods:

• Utilized a dataset of one million annual bicycle rides in The Netherlands.
• Applied various machine learning models: Random Forest, k-Nearest Neighbor, Support Vector Regression, Extreme Gradient Boosting, and Neural Networks.
• Integrated sparse GPS cycling data with external information like weather and intersection complexity.

Main Results:

• Demonstrated the feasibility of estimating bicycle delays with incomplete GPS data.
• Machine learning models successfully predicted delays by incorporating supplementary data sources.
• Showcased the value of combining sparse sensor data with contextual information.

Conclusions:

• Data-driven approaches, particularly machine learning, can effectively estimate bicycle delays.
• Sparse GPS data, when augmented with external information, is a viable resource for transportation analysis.
• Findings support informed traffic management, bicycle policy development, and infrastructure planning.