Abstract
An integration of classification methods promises good synergetic effects, since classification is the best-understood problem-solving method, and the knowledge forms of the different methods (simple, heuristic, statistical, case-comparing, set-covering and functional classification, see Chaps. 14–20), although they overlap somewhat, are not mutually convertible:
-
Statistical and heuristic knowledge resemble one another in the representation of probabilistic assessments. They differ, however, firstly in the variety of knowledge representation they provide, since heuristic classification systems offer mechanisms such as multiple hierarchies, rules with symptom combinations and exceptions, etc., which have no equivalents in statistical knowledge representation systems, and secondly in the precision of the probabilistic assessments, which are calculated from data bases or estimated by experts, respectively. The importance of the difference between calculated and estimated symptom/diagnosis probabilities was shown by experiments with a successful program by de Dombal, based on Bayes’ Theorem: the success rate of the program dropped drastically when the statistical probabilities were replaced by experts’ estimates [de Dombal 72].
-
Heuristic and set-covering classification are both based on empirical knowledge about faulty behavior of the system being investigated. They differ, however, in the direction of derivation (symptom → diagnosis and diagnosis → symptom, respectively) and also in the fact that the experts ’ estimates play an important role in heuristic systems, while set-covering systems rather check whether diagnoses are consistent with the observed symptoms.
-
Set-covering and functional classification resemble one another in the way they match observed and predicted symptoms. They differ, however, in the conceptual function of symptoms and diagnoses. In set-covering classification the diagnoses and symptoms, represented as states, are fixed concepts in their own right, as in all other types of knowledge, whereas in functional classification they are derived from the functional model: diagnoses are components with an altered input/output behavior, and symptoms are abnormal parameter values of the materials.
-
Case-comparing classification differs basically from all other knowledge types, since it is based not so much on an abstraction of case knowledge but rather on a direct comparison of a new case with those stored in a data base.
-
Simple classification can be regarded as a compiled form of the other problem-solving methods, in which the uncertainties or causal relationships are no longer explicitly represented.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Blum, R.: Discovery, Confirmation and Incorporation of Causal Relationships from a Large Time-Oriented Clinical Data Base: The RX Project, in Clancey, W. and Shortliffe, E. (eds.): Readings in Medical Artificial Intelligence: the First Decade, Chap. 17, Addison-Wesley, 1984.
Blum, R.: Modeling and Encoding Clinical Causal Relationships in a Medical Knowledge Base, in Miller, P. (ed.): Selected Topics in Medical Artificial Intelligence, Chap. 9, Springer, 1988.
Chandrasekaran, B. and Mittal, S.: Deep versus Compiled Knowledge Approaches to Diagnostic Problem Solving, AAAI-82, 349–354, 1982.
de Dombai, F., Leaper, D., Horrocks, J., Staniland, J. and McCann, A.: Computer-Aided Diagnosis of Acute Abdominal Pain, British Med. Journal 2, 9–13, 1972.
de Jong, G. and Mooney, R.: Explanation Based Learning: An Alternative View, Machine Learning 1, 145–176, 1986.
Eshelman, L.: MOLE: A Knowledge Acquisition Tool for Cover-and-Differentiate Systems, in Marcus, S. (ed.): Automating Knowledge Acquisition for Expert Systems, Chap. 3, Kluwer Academic Publishers, 1988.
Kahn, G.: MORE - From Observing Knowledge Engineers to Automating Knowledge Acquisition, in Marcus, S. (ed.): Automating Knowledge Acquisition for Expert Systems, Chap. 2, Kluwer Academic Publishers, 1988.
Kobsa, A. and Wahlster, W. (eds.): User Models in Dialog Systems, Springer, 1990.
Koton, P.: Using Experience in Learning and Problem Solving, Dissertation, Lab. for Computer Science, MIT, 1988.
Michalski, R. and Chilausky, R.: Learning by Being Told and Learning from Examples: an Experimental Comparison of Two Methods of Knowledge Acquisition in the Context of Developing an Expert System for Soyabean Disease Diagnosis, Policy Anal, and Inf. Systems 4, 125–160, 1980.
Steels, L.: Second Generation Expert Systems, Future Generation Computing Systems 1, 213–221, 1985.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1993 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Puppe, F. (1993). Integration of Classification Methods. In: Systematic Introduction to Expert Systems. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-77971-8_34
Download citation
DOI: https://doi.org/10.1007/978-3-642-77971-8_34
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-77973-2
Online ISBN: 978-3-642-77971-8
eBook Packages: Springer Book Archive