Download the full-sized PDF
Permanent link (DOI): https://doi.org/10.7939/R3BC3T540
This file is in the following communities:
|Graduate Studies and Research, Faculty of|
This file is in the following collections:
|Theses and Dissertations|
Statistically Sound Interaction Pattern Discovery from Spatial Data Open Access
- Other title
Spatial data mining
- Type of item
- Degree grantor
University of Alberta
- Author or creator
- Supervisor and department
Sander, Joerg (Computing Science)
- Examining committee member and department
Parsons, Ian (Computing Science)
Zaiane, Osmar (Computing Science)
Sanchez-Azofeifa, Arturo (Earth and Atmospheric Sciences)
Shekhar, Shashi (Computer Science - University of Minnesota)
Nascimento, Mario (Computing Science)
Department of Computing Science
- Date accepted
- Graduation date
Doctor of Philosophy
- Degree level
Spatial interaction pattern mining is the process of discovering patterns that occur due to the interaction of Boolean features from a spatial domain. A positive interaction of a subset of features generates a co-location pattern, whereas a negative interaction of a subset of features generates a segregation pattern. Finding interaction patterns is important for many application domains such as ecology, environmental science, forestry, and criminology.
Existing methods use a prevalence measure, which is mainly a frequency based measure. To mine prevalent patterns, the known methods require a user defined prevalence threshold. Deciding the right threshold value is not easy and an arbitrary threshold value may result in reporting meaningless patterns and even not reporting meaningful patterns. Due to the presence of spatial auto-correlation and feature abundance, which are not uncommon in a spatial domain, random patterns may achieve prevalence measure values higher than the used threshold just by chance, in which case the existing algorithm will report them. To overcome these limitations, we introduce a new definition of interaction patterns based on a statistical test. For the statistical test, we propose to design an appropriate null model which takes spatial auto-correlation into account. To reduce the computational cost of the statistical test, we also propose two approaches.
Existing mining algorithms also use a user provided distance threshold at which the algorithm checks for prevalent patterns. Since spatial interactions, in reality, may happen at different distances, finding the right distance threshold to mine all true patterns is not easy and a single appropriate threshold may not even exist. In the second major contribution of this thesis, we propose an algorithm to mine true co-locations at multiple distances. Our approach does not need thresholds for the prevalence measure and the interaction distance. An approximation algorithm is also proposed to prune redundant patterns that could occur in a statistical test. This algorithm finally reports a minimal set of patterns explaining all the detected co-locations. We evaluate the efficacy of our proposed approaches using synthetic and real data sets and compare our algorithms with the state-of-the-art co-location mining approach.
- Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.
- Citation for previous publication
Sajib Barua and Joerg Sander. SSCP: Mining Statistically Significant Co-location Patterns. In Proceedings of the 12th International Symposium on Advances in Spatial and Temporal Databases, pages 2–20, Minneapolis, MN, USA, 2011.Sajib Barua and Joerg Sander. Mining Statistically Significant Co-location and Segregation Patterns. Transactions on Knowledge and Data Engineering, 99 (PrePrints):1, 2013.
- Date Uploaded
- Date Modified
- Audit Status
- Audits have not yet been run on this file.
File format: pdf (Portable Document Format)
Mime type: application/pdf
File size: 1744123
Last modified: 2015:10:12 14:26:49-06:00
Filename: Barua_Sajib_Spring 2014.pdf
Original checksum: 64c2d8b0b5a409d24bf16a5594b53a33
Well formed: true
File title: Introduction
File title: Statistically Sound Interaction Pattern Discovery from Spatial Data
File author: Sajib Barua
Page count: 150