Analyzing Walking Route Choice through Built Environments using Random Forests and Discrete Choice Techniques.

Area of Science:

• Urban Planning
• Transportation Science
• Public Health

Background:

• Walking offers significant health, environmental, and community benefits.
• Understanding the built environment's influence on walking is crucial for urban design.
• Human mobility data and walkability data offer rich insights but pose analytical challenges for traditional models.

Purpose of the Study:

• To develop a novel method combining data-driven techniques with route choice modeling for walkability analysis.
• To leverage random forests for variable selection in walking route choice models.
• To compare the performance of data-driven and theory-driven route choice models.

Main Methods:

• Utilized GPS and cell phone mobility data alongside high-resolution walkability data from Salt Lake City, Utah.
• Applied the random forest technique for data-driven variable selection.
• Estimated and compared data-driven and theory-driven walking route choice models.

Main Results:

• The random forest technique effectively selected variables that significantly improved the goodness of fit for walking route choice models.
• Data-driven models demonstrated superior performance compared to models based on predefined walkability dimensions.
• Key built environmental factors influencing walking behavior were identified.

Conclusions:

• Data-driven approaches, particularly random forests, enhance the accuracy and interpretability of walking route choice models.
• This methodology provides valuable insights for urban planners and policymakers aiming to promote walking and create more walkable communities.
• Integrating advanced analytical techniques with mobility data is essential for optimizing urban design for active transportation.