This book is a monograph on practical aspects of probabilistic networks (a.k.a.
probabilistic graphical models) and is intended to provide a comprehensive
guide for practitioners that wish to understand, construct, and analyze decision
support systems based on probabilistic networks, including a number
of different variants of Bayesian networks and influence diagrams. The book
consists of three parts:
• Part I: Fundamentals of probabilistic networks, including Chapters 1–5,
covering a brief introduction to probabilistic graphical models, the basic
graph-theoretic terminology, the basic (Bayesian) probability theory, the
key concepts of (conditional) dependence and independence, the different
varieties of probabilistic networks, and methods for making inference in
these kinds of models. This part can be skipped by readers with fundamental
knowledge about probabilistic networks.
• Part II: Model construction, including Chapters 6–8, covering methods and
techniques for elicitation of model structure and parameters, a large number
of useful techniques and tricks to solve commonly recurring modeling
problems, and methods for constructing probabilistic networks automatically
from data, possibly through fusion of data and expert knowledge.
Chapters 6 and 7 offer concrete advice and techniques on issues related
to model construction, and Chapter 8 explains the theory and methods
behind learning of Bayesian networks from data.
• Part III: Model analysis, including Chapters 9–11, covering conflict analysis
for detecting conflicting pieces of evidence (observations) or evidence
that conflicts with the model, sensitivity analysis of a model both with
respect to variations of evidence and model parameters, and value of information
analysis. This part explains the theory and methods underlying
the three different kinds of analyses.