A Data-Driven Customer-Search Modeling With the Consideration of Traffic Environment

Area of Science:

• Urban transportation systems
• Behavioral modeling
• Data mining

Background:

• Understanding vacant taxi drivers' customer search behavior is crucial for optimizing urban mobility.
• Existing models often overlook the dynamic and complex factors influencing driver decision-making.

Purpose of the Study:

• To calibrate a time-dependent Multinomial Logit (MNL) model to explore determinants of vacant taxi drivers' customer-search behavior.
• To identify key factors influencing taxi drivers' search for passengers in urban environments.

Main Methods:

• Utilized over 1.6 billion GPS records from approximately 8,400 taxis in Shanghai, China.
• Applied the Ordering Points To Identify Clustering Structure (OPTICS) algorithm to divide the city into 47 hotspots.
• Employed the maximum likelihood method to identify significant factors affecting customer-search behavior.

Main Results:

• Relative passenger demand, regional pick-up likelihood, and expected rate of return are the most significant factors influencing driver search behavior.
• Traffic conditions, while less significant, still offer opportunities for service optimization and congestion mitigation.
• Customer-search behavior demonstrates significant variation depending on the time of day.

Conclusions:

• The study provides critical insights into the factors driving vacant taxi driver behavior, essential for future taxi service operations.
• Findings are applicable to mixed fleets of human-driven and self-driving taxis, informing operational strategies.
• Recommendations for service providers and policymakers to leverage identified factors for improved taxi dispatch and traffic management.