Models and Issues on Probabilistic Data Streams with Bayesian Networks
Hideyuki Kawashima, Ryo Sato, Hiroyuki Kitagawa · 2008
This paper proposes the integration of probabilistic data streams and relational database by using Bayesian networks that is one of the most famous techniques for expressing uncertain contexts. A Bayesian network is expressed by the graphical model while relational data are expressed by relation. To integrate them we make the relational model as the unified model for its simplicity. A Bayesian network is modeled as an abstract data type in an object relational database, and we define signatures to extract a probabilistic relation from a Bayesian network. We provide a scheme to integrate a probabilistic relation and normal relations. To allow continual queries over streams for a Bayesian network, we introduce a new concept, lifespan.