Spatio-temporal Mining of Ride-hailing Travel Patterns

Ride-hailing orders encode rich travel intent, but the data is sparse and unevenly distributed; appropriate clustering and visualization are needed to distill patterns.

Mine typical spatio-temporal travel patterns from ride-hailing orders to inform demand forecasting and fleet scheduling.

Mapped orders into an OD grid tensor, decomposed via NMF into base patterns; clustered hotspots with HDBSCAN and analyzed tidal characteristics over time; visualized in QGIS and ECharts.

Travel patterns mirror urban structure—commuting reveals job-housing separation, entertainment maps the night economy. Data makes the city legible.