In Partial Fulfillment of the Requirements for the Degree of
Master of Science
Will defend his thesis
In recent years the widespread usage of scanning device, such as GPS-enabled devices, PDAs, and video cameras, has resulted in an abundance of spatial data. Therefore, there is an increasing interest in mining hidden patterns in spatial data. Discovery of co-location patterns has been a research area in association analysis, pattern recognition a geographical information systems for several years.
In this thesis, we designed and implemented a user friendly, interactive Co-location Analysis Tool which can be used to extract co-location patterns from spatial datasets. By using this tool, we are able to extract co-location patterns at different levels of granularity; these results can help with business decision-making, ecology science research and urban planning. The tool provides two approaches to analyze collocation patterns: Ripley’s K-function approach and a novel approach called K-Nearest-Neighbor Distance approach. Both approaches compute spatial statistics for different neighborhood sizes and compare these characteristics with spatial characteristics obtained by placing objects randomly to determine the presence of collocation and anti-collocation. The second approach uses summaries of k-nearest neighbor distances of objects in the dataset to diagnose the presence of collocation patterns. In addition, the tool provides visualization techniques to present the data analysis experimental results. Finally, we validated the tool and compared the two collocation analysis approaches for a building dataset.
Date: Monday, June 9, 2014
Time: 2:10 PM
Place: PGH 362
Faculty, students, and the general public are invited.
Advisor: Prof. Christoph Eick