Fuzzy Rule Learner (FURL) is a the- ory revision approach to fuzzy rules learning based on Hierarchical Pri- oritized Structures. Each new level is composed from exceptions to rules from the preceding levels. The new rules are chosen in order to elimi- nate the biggest classification errors found in the training data. FURL may me combined with many tech- niques used to interpret rule bases in fuzzy controllers. In the traditional approaches to fuzzy approximation, the learning of rules has an undesirable effect. When many new rules are added, the interpretation of the rule base tends to one of its extreme values, thus we loose its informational value. In this paper, we suggest and test two methods which may overcome this drawback, negated antecedents and a controller with conditionally firing rules. We show that they allow to improve the performance of systems based on learning of fuzzy rules, namely the Fuzzy Rule Learner. The methods are tested on ECG and Multiple Sclerosis Disease datasets.
NAVARA M, PERI D (2004). Automatic Generation of Fuzzy Rules and its Applications in Medical Diagnosis. ??????? it.cilea.surplus.oa.citation.tipologie.CitationProceedings.prensentedAt ??????? Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU 04).
Automatic Generation of Fuzzy Rules and its Applications in Medical Diagnosis
PERI, Daniele
2004-01-01
Abstract
Fuzzy Rule Learner (FURL) is a the- ory revision approach to fuzzy rules learning based on Hierarchical Pri- oritized Structures. Each new level is composed from exceptions to rules from the preceding levels. The new rules are chosen in order to elimi- nate the biggest classification errors found in the training data. FURL may me combined with many tech- niques used to interpret rule bases in fuzzy controllers. In the traditional approaches to fuzzy approximation, the learning of rules has an undesirable effect. When many new rules are added, the interpretation of the rule base tends to one of its extreme values, thus we loose its informational value. In this paper, we suggest and test two methods which may overcome this drawback, negated antecedents and a controller with conditionally firing rules. We show that they allow to improve the performance of systems based on learning of fuzzy rules, namely the Fuzzy Rule Learner. The methods are tested on ECG and Multiple Sclerosis Disease datasets.File | Dimensione | Formato | |
---|---|---|---|
NavaraPeri-IPMU04.pdf
Solo gestori archvio
Dimensione
106.93 kB
Formato
Adobe PDF
|
106.93 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.