Applying Analytics

Unstructured Data Sets 183

part

correctly made o r faulty







@

@

@R

cheap test

pass or fail

exp ensive test

pass or fail

It is easy to imagine an analyst needing to answer a variety of questions

for the testing scenario I’ve described; for example:

• If a part fails the cheap test but passes the expensive test, what is the

probability that it is correctly ma de?

• If a part fails the cheap test, what is the proba bility that it will pass the

exp ensive test?

• If a par t is faulty, what is the probability that it would pass both tests?

Bayesian belief networks use the algebraic connections between the nodes

given by Bayes’ r ule to answer questions such as