dc.contributor.advisor |
Kaiqi, Zhao |
|
dc.contributor.advisor |
Jiamou, Liu |
|
dc.contributor.author |
Cheng, Wei |
|
dc.date.accessioned |
2022-09-07T23:39:03Z |
|
dc.date.available |
2022-09-07T23:39:03Z |
|
dc.date.issued |
2022 |
en |
dc.identifier.uri |
https://hdl.handle.net/2292/61074 |
|
dc.description.abstract |
Recent years have seen rapid development and proliferation of devices carrying GPS chips,
such as mobile phones. These GPS chips continuously record the timestamps and GPS
coordinates of moving objects, generating a large amount of trajectory data at an unprecedented
scale. Mining and analyzing these trajectory data provides us valuable insights
into many areas of public interest, including urban planning, transportation management,
and tourism planning. One data mining task that attracts many research interests is trajectory
anomaly detection. Particularly, for trajectory data collected from taxis, we are
interested in the anomalous trajectories as they could be caused by taxi frauds or anomalous
events. This task is not trivial since the normal routes in the real world can be very
di erent from place to place. In addition, previous studies mainly focus on utilizing the
spatial information of the trajectory to detect anomalies. In this research, we propose
a novel time-aware framework that combines both spatial and temporal information to
encode trajectories into a latent route space. We model the latent route space with a
Gaussian mixture model to capture routes of di erent types (e.g., commute during peak
hours). We use the detect-by-generation strategy for detection, which means the anomaly
score of a trajectory is computed by the likelihood that the trajectory follows a normal
route in the latent route space. Experimental results show that our time-aware framework
outperforms all baseline models, including the state-of-the-art GMVSAE model, in both
o ine and online anomaly detection tasks. We also use visualization techniques to show
that our framework captures time-aware driving patterns successfully. |
|
dc.publisher |
ResearchSpace@Auckland |
en |
dc.relation.ispartof |
Masters Thesis - University of Auckland |
en |
dc.relation.isreferencedby |
UoA |
en |
dc.rights |
Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated. |
|
dc.rights.uri |
https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm |
en |
dc.rights.uri |
http://creativecommons.org/licenses/by-nc-sa/3.0/nz/ |
|
dc.title |
Time-aware Trajectory Anomaly Detection |
|
dc.type |
Thesis |
en |
thesis.degree.discipline |
Computer Science |
|
thesis.degree.grantor |
The University of Auckland |
en |
thesis.degree.level |
Masters |
en |
dc.date.updated |
2022-08-02T05:30:30Z |
|
dc.rights.holder |
Copyright: the author |
en |
dc.rights.accessrights |
http://purl.org/eprint/accessRights/OpenAccess |
en |