University of Houston

Due to the advances in remote sensors and sensor networks, different types of dynamic and geographically distributed spatial and spatio-temporal data become increasingly available. Spatial and temporal properties are key aspects of data analysis in many applications, such as geographic information systems, weather forecasting, medical imaging, etc. The combination with spatial and temporal dimensions in large spatio-temporal data has introduced new challenges to data mining and knowledge discover. Traditional clustering techniques are inefficient in clustering spatial and spatio-temporal data because they do not incorporate the idiosyncrasies of the spatial domain. New techniques are needed to address these challenges.

The goal of this research is to develop new spatial clustering algorithm, spatio-temporal clustering algorithm, and post-analysis techniques for mining spatial and spatial-temporal datasets, especially polygons that are spatially overlapped and dynamically change their locations, sizes and shapes though time. The specific objectives are: (i) to develop density-base spatial clustering algorithms to find spatial patterns, (ii) to implement density-based spatio-temporal clustering algorithms to identify spatio-temporal patterns, (iii) to perform cluster analysis to interpret the identified patterns, and (iv) to evaluate our algorithms by example of multiple real word spatial and spatio-temporal datasets, such as datasets involving in ozone pollution events in Houston metropolitan area.