Spatial and Spatio-Temporal Clustering

Due to the advances in technology, such as smart phones, general mobile devices, remote sensors, and sensor networks, different types of spatial data become increasingly available. These data can also integrate multiple other types of information, such as temporal information, social information, and scientific measurements, which provide a tremendous potential for discovering new useful knowledge, as well as new research challenges. In this research, we focus on clustering and analyzing spatial and spatio-temporal data. We have addressed several important sub-problems in polygon-based spatial and spatio-temporal clustering and post-processing analysis techniques. We have developed (1) two distance functions that measure the distances between polygons, especially overlapping polygons; (2) a density-based spatial clustering algorithm for polygons; (3) two post-processing analysis techniques to extract interesting patterns and useful knowledge from spatial clusters; (4) two density-based spatio-temporal clustering algorithms for polygons; (5) a box plot based post-processing analysis technique to identify interesting spatio-temporal clusters of polygons; (6) a change-pattern-discovery algorithm to detect and analyze patterns of dynamic changes within spatio-temporal clusters of polygons; and (7) a formal definition of the task of finding uniform regions in spatial data and an algorithm to identify such uniform regions. Our algorithms and techniques are demonstrated and evaluated in challenging real-world case studies involving ozone pollution events in the Houston-Galveston-Brazoria area and the building data of Strasbourg, France. The results show that our algorithms are effective in finding compact clusters in spatial and spatio-temporal domains and in extracting interesting patterns and useful information from spatial and spatio-temporal data.