Incremental update of spatial association patterns
Abstract
Spatial data mining is the process of discovering interesting and useful patterns from a large spatial dataset, where spatial data means data pertaining to geographic references. As one of the important spatial data mining tasks, spatial association pattern mining has been popularly studied in spatial data mining literature. As one of the spatial association patterns, colocation patterns represent subsets of spatial events which are frequently observed together in a nearby area. This thesis explores the problem of computing colocation patterns in evolving spatial databases. The databases in environmental monitoring, transportation or mobile application domains tend to be very dynamic, e.g., constantly updated with fresh data and removed data. Changes to the data can invalidate current colocation patterns or introduce new patterns. When a large spatial database is updated, it is nontrivial to maintain colocation patterns current since new data objects can make neighbor relationships with existing data objects as well as other new data objects in continuous space. This thesis addresses the case that arises after new data points have been added to the spatial database, and proposes an algorithm for efficiently updating spatial colocation patterns. This work also addresses how the proposed algorithm can be extended for the case that arises after a nontrivial number of existing data points have been deleted from the spatial database. The experimental results show that the proposed algorithm is efficient in the computation of updating colocation patterns.
Degree
M.S.
Advisors
Yoo, Purdue University.
Subject Area
Information Technology|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.