|
Abstract:
|
We address the problem of estimating count queries on databases quickly, without accessing the database at query time. We accomplish that we build a model of the domain from the database in a preprocessing phase and use this to answer count queries. The model we use is the Mixture of Bayesian Networks (MBNs) , which effectively encodes the joint probability distribution of the domain. An MBN is a weighted model with Bayesian networks as components. We describe how to learn an MBN model from a database using an instance of the modified Expectation-Maximization (EM) algorithm, called the EAM algorithm, and evaluate its accuracy on real and artificial data sets. Experimental results show that MBNs can represent a data set satisfactorily and can approximate counts with a high degree of accuracy, without accessing the database. We illustrate our method by applying it on a biological database.
|
