Exploring ridesourcing trip patterns by fusing multi-source data: A big data approach

Abstract

Ridesourcing service has been playing an increasingly important role in urban travel, which it can be regard as a promising path towards urban sustainability because of its nature of “sharing”. This paper investigated the travel behavior of ridesourcing users, by considering pick-up and drop-off locations fusing Didi ridesourcing data and online Point of Interest (POI) data based on the Latent Dirichlet Allocation (LDA) model. A case study was conducted in Chengdu, an Asian city. Firstly, the study area was tessellated with hexagons and then clustered into five hexagon types based on POI data. Then researchers analyzed information on spatial distribution of POI data in five hexagon types and temporal distribution of boarding and alighting ridership for five hexagon types. The validation results indicated that the expression “Nine to Ten” represents the mode of life for people in Chengdu. Ridesourcing is not yet being used as a commuter tool for most people. LDA model results also verified that many people work overtime in the evenings and especially on Saturday, even though Chengdu is often thought of as a “slow city”. The study will be useful for practitioners and government to implement effective policies about multi-modal transportation.