Representing And Real-time Probabilistic Query Processing To Uncertain Data With Complex Correlation

Numerous applications which involve uncertain data and their complex correlation(including spatio/temporal correlations), eg., sensor network, RFID, market prediction, information extraction, etc., brought great challenges to the database theory and technology in managing and querying large amounts of this complex uncertain data. Most probabilities database models proposed to date, however, fails to meet the challenges. Some are often too simple to represent more complex correlations effectively. Although most others relies on graphical modeling techniques (eg., Baysesian networks) and can capture the representation of complex uncertainty and correlations, the scalability of query and probabilistic reason cannot satisfy the needs of the application.In this work, a novel probabilistic data management model RTx-PDM is proposed.RTx-PDM can not only handle arbitrary probabilistic uncertain data natively at the attribute or tuple level but also represent the correlation patterns between uncertain data with the intuitive BLOCK structure. The probabilistic attributes of tuples in the same BLOCK have the same correlation patterns, therefore, they support to share the correlation patterns in an efficient manner. As the BLOCK structure bears the hierarchy, nesting, and inherited trait, the uncertain data correlation patterns can be represented more flexibly and efficiently. Besides, the RTx-PDM's completeness for query processing in probabilistic database is proved.In this thesis, we propose a probabilistic query famework and also investigate various real-time probabilistic query stategy to meet the need of more important and less resource queries. Besides, we also propose shortly and lazy reasoning arithmetic.Finally, we validate our approach by demonstrating the value of exploiting data correlations during query processing.