Doctor of Philosophy (PhD)
The next generation of database management and computing systems will be significantly complex with data distributed both in functionality and operation. The complexity arises, at least in part, due to data types involved and types of information request rendered by the database user. Time sequence databases are generated in many practical applications. Detecting similar sequences and subsequences within these databases is an important research area and has generated lot of interest recently. Previous studies in this area have concentrated on calculating similitude between (sub)sequences of equal sizes. The question of unequal sized (sub)sequence comparison to report similitude has been an open problem for some time. The problem is an important and non-trivial one. In this dissertation, we propose a solution to the problem of finding sequences, in a database of unequal sized sequences, that are similar to a given query sequence. A paradigm to search pairs of similar, equal and unequal sized, subsequences within a pair of sequences is also presented. We put forward new approaches for sequence time-scale reduction, feature aggregation and object recognition. To make the search of similar sequences efficient, we propose an indexing technique to index the unequal-sized sequence database. We also introduce a unique indexing technique to index identified subsequences within a reference sequence. This index is subsequently employed to report similar pairs of subsequences, when presented with a query sequence. We present several experimental results and also compare the proposed framework with previous work in this area.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Dua, Sumeet, "Techniques to explore time-related correlation in large datasets" (2002). LSU Doctoral Dissertations. 2981.
S. S. Iyengar