Understanding ride-sourcing drivers' working patterns based on platform operations data

Ride-sourcing has become an important part of urban transportation systems. Understanding ride-sourcing drivers' working patterns is significant for analyzing urban mobility and modeling the ride-sourcing market. This paper explores ride-sourcing drivers' working patterns using real-world trip order and vehicle trajectory big data in Ningbo, China. First, the definition and rules for identification of 'shift' are introduced, and shift features are explored. Subsequently, a two-stage clustering method is proposed to establish connections between drivers' working patterns and raw shift records. In the first stage, the Gaussian mixture model is used to divide all shift records into ten types based on their start and end times, leading to the identification of ten basic shift patterns. In the second stage, each individual driver is represented using a vector indicating his/her visit frequencies over the ten basic shift patterns, and the K-Means model is used to divide all drivers into seven groups with distinct shift preferences. The results uncover the heterogeneity and complexity of drivers' working patterns. Furthermore, the analysis sheds light on drivers' circumstances, enriching our understanding of their working behavior. Finally, several managerial implications for the government and the platform are given according to the findings. The managerial implications for government policymakers and platform operators can support them in decisions about driver wage guarantee, fleet size restriction, market condition feedback, and new service development.