Comparison of spatial data indexing techniques

The diploma thesis deals with the comparison of different approaches to indexing spatial data in DBMS. A representative sample of database management systems consists of MySQL, SQLite and PostgreSQL, respectively their various implementations of the spatial indexes R-tree, R*-tree, GiST and SP-GiST. Individual approaches are evaluated on the basis of open data sets originating from the Národní katalog otevřených dat (NKOD) and real estate data from the Sreality advertising portal, both in terms of computational complexity. The work is divided into a part summarizing theoretical knowledge and procedures and a practical part, in which these knowledge and procedures are used. The first part presents the definition, requisites and functions of geographic information system and spatial database systems, methods of representation of multidimensional objects and general principles of operation of these indexes and database operations performed on them. The practical part captures the acquisition of comparable data using benchmarking of spatial operations, especially spatial relations, based on the DE-9IM model, utility functions and special algorithms, including the nearest neighbor search, using a prepared sample database aggregating various data sets. The output is used test scripts and an overview matrix classified according to defined criteria and measured values for each DBMS, which can be the basis for the reader to select the ideal database management system, respectively the type of index for the use case the reader requires.