PGMC: a framework for probabilistic graphic model combination

Publication Type

Conference Proceeding Article

Publication Date

10-2005

Abstract

Decision making in biomedicine often involves incorporating new evidences into existing or working models reflecting the decision problems at hand. We propose a new framework that facilitates effective and incremental integration of multiple probabilistic graphical models. The proposed framework aims to minimize time and effort required to customize and extend the original models through preserving the conditional independence relationships inherent in two types of probabilistic graphical models: Bayesian networks and influence diagrams. We present a four-step algorithm to systematically combine the qualitative and the quantitative parts of the different models; we also describe three heuristic methods for target variable generation to reduce the complexity of the integrated models. Preliminary results from a case study in heart disease diagnosis demonstrate the feasibility and potential for applying the proposed framework in real applications.

Keywords

Algorithm; article, Artificial neural network, Bayes theorem, Decision support system, Feasibility study, Heart disease, Human, Statistical model

Discipline

Computer Sciences | Health Information Technology

Publication

AMIA 2005 Symposium Proceedings

Volume

2005

First Page

370

Last Page

374

Publisher

AMIA

City or Country

Washington DC, USA

This document is currently not available here.

Share

COinS