Skip to Main content Skip to Navigation
Theses

Floating Car Data mining to feature out mobility patterns : individual-centered and place-based analyses

Abstract : The presence and movement of human beings in space and time constitute their mobility: it is a physical phenomenon and also a socioeconomic phenomenon, since people choose their locations and their trips between them to satisfy their needs and desires. The scientific knowledge of mobility as a physical phenomenon involves causalities and patterns. Owing to the recent surge in sensing technologies, trajectory data are now massively available, thereby empowering the observation of human mobility. Among various sources, Floating Car Data (FCD) pertain to vehicle-based mobility and yield discretized trajectories of positions in space and time for the “vehicle” entity. Most of the academic literature has concentrated on methods for characterizing traffic conditions from an engineering aspect. Only a few studies have shifted the focus of FCD data mining towards semantic-oriented excavation by exploring behavioral representations.This thesis aims to explore and analyze mobility patterns by leveraging FCD to contribute a better understanding of vehicle-based movements. More specifically, mobility patterns are studied at two levels: the individual level of human behaviors from the trajectory “authors”, and the more global level of capturing spatial relations and structure on the basis of aggregated mobility features in space. Furthermore, another fold of the thesis objective is to build up methodological approaches for trajectory mining, contributing to broadening the way of using trajectory data in mobility analytics.The first part of the thesis pertains to individual mobility with three research questions addressed. First, Chapter 2 discovers vehicle usage patterns based on their daily mobility-making and constitute a typology for the ways of vehicle usage. Then, Chapter 3 investigates significant places to mobility makers, by identifying the “anchoring” geolocations and further extrapolating their meaningful representations such as homes, workplaces, and other secondary places. Next, Chapter 4 proposes a novel approach for travel time estimation by building a stochastic model to exploit FCD materials.The second part of the thesis pertains to places and spatial relations with another three research questions addressed. First, Chapter 5 aims to reveal the functional occupation of urban areas by looking at related vehicle movements and further characterize the spatial land use divisions. Second, Chapter 6 studies the spatial organization of a territory, with a particular emphasis on the jobs-housing spatial relations. It establishes a data-driven method to recognize employment core areas by density and identify corresponding residential catchment areas by core-periphery patterns. Third, Chapter 7 investigates the spatial interaction between places. It deals with the estimation of the Origin-Destination matrix flows based on two kinds of data: vehicle trajectory data and local traffic counts, along with a Bayesian assignment framework to account for the heterogeneous sampling rate issues of such data.Overall, this thesis expands the “mobility analytics” of “mobility patterns” from a data mining standing point. It contributes to overcoming traditional limitations on extensive mobility analysis in terms of inter-day variations and large-scale observations by employing massive digital trajectories and artificial intelligence. Through various applications, this thesis shows the feasibility of mining semantic context behind individual mobility at a micro-level and the possibility of capturing grouped phenomena reflected in geographical spaces at a macro-level. However, this thesis pays particular attention to vehicle-related mobility based on FCD. Future work can bring with other modes of transportation to have a more complete investigation of the mobility system.
Document type :
Theses
Complete list of metadata

https://pastel.archives-ouvertes.fr/tel-03637330
Contributor : ABES STAR :  Contact
Submitted on : Monday, April 11, 2022 - 3:29:26 PM
Last modification on : Friday, April 15, 2022 - 2:50:47 PM
Long-term archiving on: : Tuesday, July 12, 2022 - 6:41:28 PM

File

TH2022ENPC0004.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03637330, version 1

Collections

Citation

Danyang Sun. Floating Car Data mining to feature out mobility patterns : individual-centered and place-based analyses. Geography. École des Ponts ParisTech, 2022. English. ⟨NNT : 2022ENPC0004⟩. ⟨tel-03637330⟩

Share

Metrics

Record views

101

Files downloads

43