In this thesis, I propose an indexing data structure within a relational system to index multidimensional data. The index is used to store and retrieve information about regions called Minimum Bounding Rectangles. The technique is used as a preprocessing step in similarity searches over time-series data. The structure is multi-resolutional, storing information with respect to different window sizes. I define a mapping of the index to relational tables, which are then indexed with B+-tree indexes.; I incorporate declarative language queries in the retrieval process to achieve sequential access patterns instead of random. In particular, I propose two SQL-based procedures for the index retrieval process: PrefixSearch and MultiMatching. I investigate the query plans issued by the DBMS optimizer. The proposed technique is compared against other existing solutions with respect to very large datasets. I also consider scalability issues and possible future extensions.; Keywords: time-series databases, relational databases, similarity searches, SQL... |