Support for Diagnosis: Introduction
From DSL
Diagnosis is one of the most successful applications of Bayesian networks. The ability of probabilistic knowledge representation techniques to perform a mixture of both predictive and diagnostic inference makes it very suitable for diagnosis. Bayesian networks can perform fusion of observations such as predispositions and risk factors with symptoms and test results. GeNIe 2 now incorporates all the diagnostic features of the 'GeNIe Diagnosis' software. This section reviews these special diagnostic features of the GeNIe 2.
A diagnostic model built using GeNIe represents various components of a system, possible faulty behaviors produced by system (symptoms), along with results of possible diagnostic tests. The model essentially captures how possible defects of a system (whether it is a human-made device, such as a car, an airplane, or a copier) can manifest themselves by error messages, symptoms, and test results. Using such a model, GeNIe produces a ranked list of the most likely defects and a ranked list of the most informative and cost-effective tests. The following section assumes that you are already well versed in using the plain version of GeNIe. If you are not, please go through the tutorials to learn more.
This chapter is structured as follows. First, in section Enabling Diagnosis, we will see how to enable the diagnostic features of GeNIe and discuss the values that properties of nodes should take to perform diagnosis on them.
Section Spreadsheet View discusses a special extension of GeNIe that is useful in rapid model building - all properties of every variable are listed in one window and the user specifying a model can move rapidly between variables and enter their specifications into the model.
Section Diagnostic window describes a special dialog window that allows the user to test the diagnostic model on real cases. The window allows for observing symptoms and signs, enter test results, and see GeNIe's suggestions as to what tests to perform next and what the probabilities of various faults are. These are listed rank-ordered according to their probability.
Finally, the Diagnostic case management section discusses how diagnostic cases can be saved to and retrieved from permanent storage (disk files).
