Date of Award

5-2011

Degree Type

Masters Thesis

Degree Name

Master of Science (MS)

Department

Computing

Committee Chair

Louise Perkins

Committee Chair Department

Computing

Abstract

A Bayesian Network is a stochastic graphical model that can be used to maintain and propagate conditional probability tables among its nodes. Here, we use a Bayesian Network to model results from a numerical riverine model. We develop an discretization optimization algorithm that improves efficiency and concurrently increases the overall accuracy of the resulting network. We measure accuracy using a new prediction accuracy criteria that includes an a posteriori soft correction. Furthermore, we show that this accuracy quickly asymptotes and begins to show diminishing returns on large data sets.

Share

COinS