An Improved Probabilistic Database Model And Its Probabilisticn Earest Neighbors Query Research

Uncertain data are more and more popular and important in the modern applications. Although the relational model for databases provides a great range of advantages, such as access flexibility, logical and physical data independence, it does not have a comprehensive way to handle incomplete and uncertain data. In 1996, Dev and Sarkalar proposed a new probabilistic data schema and probabilistic database model to handle such data, but the schema and model have many shortcomings. Thus, we must extend the relational model, handle the uncertain data so as to describe the dynamic world and establish a new probabilistic data model—PDS(Probabilistic Data Schema).We incorporate object attributes, static attributes, dynamic attributes and probabilisic attributes into PDS and propose a data theory which all of these attributes construct tuples. On the basis of the model, we define some tuple calculus. Thus we have PDS databases system and the system have the ability of handle probabilistic data.Probabilistic databases can store moving objects. R-tree is the most practical index in probabilistic databases, but the performance of query and update on R-tree is very low. PRI-Tree(Probabilistic R-Tree Index), which is based on R-tree, supplemented by a hash index on leaf nodes is presented for solving the problem. Performance analysis show that the PRI-tree update performance outperforms that on R-Tree.Uncertainty is commonly seen in the sensor database. So how to explicitly query uncertain data is the key problem. Probabilistic nearest neighbor query can explicitly describe uncertain data. This thesis proposes two algorithms: a Branch-Bound based algorithm and a Delaunay Tnangulation based algorithm, and presentes the performance analysis respectively.