Mining periodicity and object relationship in spatial and temporal data

Li, Zhenhui

Permalink

https://hdl.handle.net/2142/42370 Copy

Description

Title

Mining periodicity and object relationship in spatial and temporal data

Author(s)

Li, Zhenhui

Issue Date

2013-02-03T19:36:37Z

Director of Research (if dissertation) or Advisor (if thesis)

Han, Jiawei

Doctoral Committee Chair(s)

Han, Jiawei

Committee Member(s)

• Zhai, ChengXiang
• Abdelzaher, Tarek F.
• Wu, Kun-Lung

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Data mining
• spatial and temporal data
• movements
• periodicity
• object relationship
• MoveMine

Abstract

With the rapid development of positioning technologies, sensor networks, and online social media, spatiotemporal data is now widely collected from smartphones carried by people, sensor tags attached to animals, GPS tracking systems on cars and airplanes, RFID tags on merchandise, and location-based services offered by social media. While such tracking systems act as real-time monitoring platforms, analyzing spatiotemporal data generated from these systems frames many research problems and high-impact applications. During my PhD study, I have extensively studied data mining algorithms for moving objects. I have contributed several key algorithms to this exciting field. I have proposed the very first work to detect periodicity from movement data even if the movement only has rough periodicity and has lots of non-periodic random short-trajectories. I have also systematically studied a broad range of relationship patterns among moving objects in practical scenarios. Objects forming a social cluster, for example, can be efficiently extracted from large-size moving object pool even if the objects in a group only have sporadic interactions. I have further conducted an examination on when, where and how moving objects interact in a sporadic way, in order to discover semantic relationships, such as friends, colleagues and family. The algorithms have been integrated into our MoveMine system, an online system allowing people to test our data mining methods on a diverse collection of real movement datasets. For future work, my long-term research plan is to study cyber-physical systems, such as ecological systems, patient-care systems, and transportation systems. Such systems consist of a large number of interacting spatial, temporal and information components.

Graduation Semester

2012-12

Permalink

http://hdl.handle.net/2142/42370

Copyright and License Information

Copyright 2012 Zhenhui Li

Owning Collections