# Graphical Models: Methods for Data Analysis and Mining. (Book Reviews)

By Carey, Vincent J. | Journal of the American Statistical Association, March 2003 | Go to article overview

# Graphical Models: Methods for Data Analysis and Mining. (Book Reviews)

Carey, Vincent J., Journal of the American Statistical Association

Christian BORGELT and Rudolf KRUSE. New York: Wiley, 2002. ISBN 0-470-84337-3. viii+ 358 pp. \$79.95 (H).

In their prefatory overview to the volume Applications of Uncertainty Formalisms, Parsons and Hunter (1998) distinguished the three main numerical approaches to approximate reasoning under conditions of uncertainty: probability theory (e.g., Lindley 1975), evidence theory (Dempster 1968), and possibility theory (Zadeh 1978). Graphical Models: Methods for Data Analysis and Mining is a fairly thorough overview of the theory of graphical models, with a focus on the use of possibility measures to characterize uncertainty in the building and interpretation of these models.

The first chapter provides definitions of knowledge discovery in databases (KDD) and data mining, with brief reference lists and URLs for catalogs of related tools. In the second chapter, a five-page sketch of relational algebra and conditional probability is followed by 30 pages on possibility theory and the context model. This includes several simple examples, conducted with special dice, to compare the inferences achievable using probability and possibility measures. [For readers who have not encountered possibility theory, a key characterization is [PI](A U B) = max([PI](A), [PI](B)) and [PI](A ** B) = min({PI](A), [PI](B)), where A and B are any fuzzy subsets of a fuzzy universe governed by a possibility measure [PI].] This chapter is a reasonable mix of formalism and illustrative simple examples.

Chapter 3, "Decomposition," includes a fairly formal treatment of relational algebra sufficient to define the minimal decomposition of a relation. Relational algebra concepts are then connected with binary possibility measures to construct concepts of conditional possibility and conditional relational independence that are analogous to central concepts of probability theory for graphical models. Conditional probability and factorization of probability models are defined, and general possibilistic decompositions are developed. The chapter concludes with a section titled "Possibility versus Probability" that includes the following:

In other words, a marginal probability distribution states "the probability that attribute A has value a is p." This probability is aggregated over all values of all other attributes and thus refers to a one element vector (a). A marginal possibility distribution instead states "the degree of possibility of a value vector with the highest degree of possibility of all tuples in which attribute A has value a is p." That is, it refers to a value vector over all attributes of the universe of discourse, although the values of all attributes other than A are left implicit.

This is perhaps the most focused statement on the contrast between the possibilistic and probabilistic frameworks found in the book. It is followed by discussion of a number of caveats for the possibilistic approach.

Chapter 4 provides a formal treatment of graphical representation of statistical models, including the graphoid axioms, proof (deferred to an Appendix) that conditional possibilistic independence satisfies the semigraphoid axioms, graph theory, Markov properties, evidence propagation, and join-tree propagation.

Chapter 5 is a mainly technical treatment of computing projections of relations and marginalizations of distributions The EM algorithm and some accelerations are presented. Examples are reviewed, but the authors note that they "could not get hold of any real-world dataset containing 'true' imprecise attribute values, i.e., datasets with cases in which for an attribute A a set S [member of](A) with \S\ I and S [not equal to](A) was possible." In other words, the authors had access only to datasets in which imprecision coincided with total missingoess. Public datasets with censored variables are available to remedy this gap.

Chapter 6, "Naive Classifiers," describes the naive Bayes classifier (a graphical model with star-like structure) and its application to the iris dataset. …

If you are trying to select text to create highlights or citations, remember that you must now click or tap on the first word, and then click or tap on the last word.
One moment ...
Default project is now your active project.
Project items

Highlights (0)
Some of your highlights are legacy items.
Citations (0)
Some of your citations are legacy items.
Notes (0)
Bookmarks (0)

Project items include:
• Saved book/article
• Highlights
• Quotes/citations
• Notes
• Bookmarks
Notes

#### Cited article

Style
Citations are available only to our active members.

(Einhorn, 1992, p. 25)

(Einhorn 25)

1. Lois J. Einhorn, Abraham Lincoln, the Orator: Penetrating the Lincoln Legend (Westport, CT: Greenwood Press, 1992), 25, http://www.questia.com/read/27419298.

#### Cited article

Graphical Models: Methods for Data Analysis and Mining. (Book Reviews)
Settings

#### Settings

Typeface
Text size Reset View mode
Search within

Look up

#### Look up a word

• Dictionary
• Thesaurus
Please submit a word or phrase above.

Why can't I print more than one page at a time?

Help
Full screen

### How to highlight and cite specific passages

1. Click or tap the first word you want to select.
2. Click or tap the last word you want to select, and you’ll see everything in between get selected.
3. You’ll then get a menu of options like creating a highlight or a citation from that passage of text.

## Cited passage

Style
Citations are available only to our active members.

"Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn, 1992, p. 25).

"Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn 25)

"Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences."1

1. Lois J. Einhorn, Abraham Lincoln, the Orator: Penetrating the Lincoln Legend (Westport, CT: Greenwood Press, 1992), 25, http://www.questia.com/read/27419298.

## New feature

It is estimated that 1 in 10 people have dyslexia, and in an effort to make Questia easier to use for those people, we have added a new choice of font to the Reader. That font is called OpenDyslexic, and has been designed to help with some of the symptoms of dyslexia. For more information on this font, please visit OpenDyslexic.org.

To use OpenDyslexic, choose it from the Typeface list in Font settings.

## Thanks for trying Questia!

Please continue trying out our research tools, but please note, full functionality is available only to our active members.

Your work will be lost once you leave this Web page.