Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Trajectory anomaly detection and reference set algorithms - a comparative study

Sandblom, Svante LU (2024) In Master's Theses in Mathematical Sciences FMSM01 20241
Mathematical Statistics
Abstract
As society's capacity for data collection increases, the need for automated processing techniques becomes increasingly relevant. One way to evaluate data is to find the subset which does not conform to normal behaviour, i.e., anomalies. If the data is in the form of trajectories in some space, the algorithms used to find these are trajectory anomaly detection algorithms. In this paper, the trajectory anomaly detection algorithms DBTOD, TRAOD, Hierarchical clustering and TOP-EYE were implemented and evaluated on simulated data in Euclidean space. Furthermore, faster variants of the TRAOD and DBTOD algorithms using a subset of the data as a reference set were created, evaluated and then tested on the Microsoft GeoLife trajectory dataset. All... (More)
As society's capacity for data collection increases, the need for automated processing techniques becomes increasingly relevant. One way to evaluate data is to find the subset which does not conform to normal behaviour, i.e., anomalies. If the data is in the form of trajectories in some space, the algorithms used to find these are trajectory anomaly detection algorithms. In this paper, the trajectory anomaly detection algorithms DBTOD, TRAOD, Hierarchical clustering and TOP-EYE were implemented and evaluated on simulated data in Euclidean space. Furthermore, faster variants of the TRAOD and DBTOD algorithms using a subset of the data as a reference set were created, evaluated and then tested on the Microsoft GeoLife trajectory dataset. All in all, the tests demonstrated the efficiency of the DBTOD and Hierarchical clustering algorithms, including the reference set variants. It also showed how the reference set algorithms could be used on relatively smaller subsets as the data size grows, although the tests could not conclusively show that these algorithms had a lower time complexity. (Less)
Please use this url to cite or link to this publication:
author
Sandblom, Svante LU
supervisor
organization
course
FMSM01 20241
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Anomaly detection, Algorithms, Trajectory, TRAOD, DBTOD
publication/series
Master's Theses in Mathematical Sciences
report number
LUTFMS-3550-2024
ISSN
1404-6342
other publication id
2024:E72
language
English
id
9176995
date added to LUP
2024-11-01 11:24:48
date last changed
2025-02-05 13:15:47
@misc{9176995,
  abstract     = {{As society's capacity for data collection increases, the need for automated processing techniques becomes increasingly relevant. One way to evaluate data is to find the subset which does not conform to normal behaviour, i.e., anomalies. If the data is in the form of trajectories in some space, the algorithms used to find these are trajectory anomaly detection algorithms. In this paper, the trajectory anomaly detection algorithms DBTOD, TRAOD, Hierarchical clustering and TOP-EYE were implemented and evaluated on simulated data in Euclidean space. Furthermore, faster variants of the TRAOD and DBTOD algorithms using a subset of the data as a reference set were created, evaluated and then tested on the Microsoft GeoLife trajectory dataset. All in all, the tests demonstrated the efficiency of the DBTOD and Hierarchical clustering algorithms, including the reference set variants. It also showed how the reference set algorithms could be used on relatively smaller subsets as the data size grows, although the tests could not conclusively show that these algorithms had a lower time complexity.}},
  author       = {{Sandblom, Svante}},
  issn         = {{1404-6342}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master's Theses in Mathematical Sciences}},
  title        = {{Trajectory anomaly detection and reference set algorithms - a comparative study}},
  year         = {{2024}},
}