Overview: For an overview of Bayes nets and decision nets, there are two special journal issues that are suitable. The first is the winter 1991 edition of AI Magazine, which contains both “Bayesian Networks Without Tears” (Charniak91) and “Decision Analysis and Expert Systems” (Henrion&BH91). The other is volume 38 of the Communications of the ACM, which is devoted to Bayes nets and decision nets (Heckerman&MW95), and has introductory material and descriptions of real applications.
Fundamentals: In 1988 Judea Pearl wrote Probabilistic Reasoning in Intelligent Systems (Pearl88), which was the most influential and widely cited book in the formative development of Bayes nets. It is an excellent foundational book, but it doesn’t contain the latest developments, and has a very mathematical and theoretical orientation. Bayesian Networks and Decision Graphs (Jensen07) has somewhat more recent results, although the earlier Jensen96 may be even more useful to a beginner who is just interested in understanding and applying Bayes nets in a basic way, if it can be obtained.
The "bible" on Bayes nets right now is Probabilistic Graphical Models (Koller&F09). It is very comprehensive and written at a high academic level by two of the world's leading researchers on Bayes nets. In the same vein, but not as comprehensive (although covering some topics, such as inference, with greater depth) is Modeling and Reasoning with Bayesian Networks (Darwiche09).
Applied: The best book on the theory of how to apply Bayes nets to the real world is Bayesian Networks and Influence Diagrams (Kjaerulff&Madsen08). Bayesian Artificial Intelligence (Korb&Nicholson10) is more readable and introductory, and contains some additional material to construct models. A sampling of real-world applications can be found in Bayesian Networks: A Practical Guide to Applications (Pourret&Naim&Marcot08).
For the usage of Bayes nets in particular fields, see:
- Probabilistic Methods for Bioinformatics (Neapolitan09)
- Probabilistic Methods for Financial and Marketing Informatics (Neapolitan07)
- Ecology: Special Issue of Canadian Journal of Forest Research (NRC06)
- Planning Improvements in Natural Resource Management (Cain01)
- Operational Risk: Measurement and Modelling (King01)
Russell&N09 is an excellent introductory textbook for artificial intelligence that does a good job of describing Bayes nets, decision nets, dynamic decision nets, policy iteration, etc. within the overall context of intelligent agents and AI software.
Causality: A very accessible and easy (but worthwhile) read on causal models is Causal Models: How People Think About the World and Its Alternatives (Sloman05). If you are building causal Bayes nets, and haven't had a lot of experience with causal modeling, you should definitely read that book. For a more academic and advanced study on causality, read Causality: Models, Reasoning, and Inference (Pearl09). It is by the world's leading researcher on causality, and although it is at a very advanced level for the subject, you can selectively read it to get the most important ideas without much difficulty. The excellent epilogue can be read with no background knowledge required; read it first. For causality as applied to psychology, see The Mind's Arrows: Bayes Nets and Graphical Causal Models in Psychology (Glymour01).
much of the literature on Bayes nets is published within the artificial
intelligence (AI) community, it is really the combined work of the statistics
/ probability, decision analysis, operations research, and AI communities
(and increasingly other communities that are applying Bayes nets, such
as resource management, ecology, finance, medicine, etc.). Many of the
researchers who developed the theory of Bayes nets and decision nets attended
the annual “Uncertainty in Artificial Intelligence” conference from 1985
to present, and so advanced material can be found in the conference proceedings
(available from the "Conference Proceedings" section of the