Spatio-Temporal Partitioning and Location Prediction in GPS Trajectory Data
Loading...
Citations
Altmetric:
Altmetric::
Download
Associated Records
Authors
File Licence
Editors
Date
2019
Source Title
Journal Title
Journal ISSN
Volume Title
Publisher
Melbourne Authors
Affiliation
Computing and Information Systems
Document Type
PhD thesis
Subject Code
Citation
Access Status
Open Access
DOI
Access URL
Open Access URL
Open Access at PMC
Linked Resource URL
NHMRC Grant Code
ARC Grant Code
Description
© 2019 Morteza Mousavi Barroudi
Abstract
In recent years, a large volume of GPS trajectory data of the mobile objects is collected by many individuals, companies, and organizations. Various applications are being developed to extract useful information from the collected trajectory data through data analysis. Having adequate information about the position and the geometry of the Points of Interest (POIs) of the mobile objects is a key requirement in a wide range of such applications. The information about the location and the geometries of the POIs is not publicly available in various environments due to the nature of the environments (e.g. animals migration monitoring), or because of military and security issues. The main research problem of this thesis is to estimate the geometries of the POIs of the mobile objects by merely using the data extracted from GPS trajectories. In this dissertation, the geometry of a POI is called a Geometry of Interest (GOI). The methodology of estimating the GOIs is composed of five phases: (i) Stay region extraction, (ii) Clustering the stay regions aiming to estimation of geometries of the destination regions, (iii) Data aggregation, (iv) Outlier detection, and (v) Partitioning the trajectory area into a grid with inhomogeneous cells containing the GOIs. In each of the phases of the spatio-temporal partitioning process, new methods and algorithms have been proposed and evaluated based on the field data. The evaluation is performed based on the geometric similarity comparison between the geometries of the real-world POIs (ground truth) and the estimated GOIs. The experimental results prove that the proposed multi-step framework is able to estimate the GOIs with a high geometric quality. Moreover, to be able to evaluate the quality of the estimated GOIs when the ground truth is not available, two novel GOI quality measurement metrics are proposed and evaluated. The experiments show that the proposed metrics can effectively analyse the quality of the GOIs without having access to the ground truth. In this thesis, we also evaluate the effect of the quality of the estimated GOIs on the accuracy of spatio-temporal location prediction applications. Moreover, to tackle with the uncertainty of the estimated GOIs which have a negative impact on the accuracy of the spatio-temporal location prediction, a novel method is proposed which performs approximate GOI matching instead of exact GOIs matching in the sequential pattern mining of the spatio-temporal location prediction process. Evaluation results show that the proposed approximate spatio-temporal location prediction method outperforms the competitor methods in terms of the rate of true prediction and reducing the temporal prediction error.
Notes
Terms of Use
Keywords
Geomtery of Interest, GOI, POI, GPS Trajectory, Spatio-Temporal Location Prediction