Spatial Data Mining Using Branch Grafted R-tree.

Spatial data mining is a process of extraction of implicit information, such as weather patterns around latitudes, spatial features in a region, et., with a goal of knowledge discovery. The existing spatial data mining methods typically identify a specific datamining task for knowledge discovery. An example of a mining task may involve finding weather patterns in the northwestern region of U.S.A. To find such weather patterns one could employ an existing data structure, such as a B+ tree followed by the analysis of the mined weather data for knowledge discovery. This is a typical top-down approach of identifying a task, selecting a data structure, followed queries and analysis. This thesis provides a method and a simulation for mining spatial rules for the purpose of knowledge discovery. The thesis takes a bottom up approach: it employs Branch Grafted R-tree for the storage and retrieval of spatial data, followed by identifying tasks, followed by spatial queries and analysis. The Branch Grafted R-tree is an efficient data structure more suitable for efficient retrieval of data. This type of bottom up approach is unique and takes the advantage of the previous work carried out using Branch Grafted R-tree.