The implementation of a multiple classifier system implies the definition of a rule (combining rule) for determining the most likely class, on the basis of the class attributed by each single classifier. The availability of a criterion to evaluate the reliability of the decision taken by a classifier can be profitably used in order to implement an effective combining rule. In this paper, we propose a method that evaluates the reliability of each classification act by using an e-Support Vector Regression approach. This idea yields to define four combining rules that work also with classifiers providing as their only output the guess class. The results obtained on some standard datasets by these reliability-based rules are compared with those obtained by using different well-known combining criteria, in order to assess the effectiveness of the proposed approach.
Evaluating Classification Reliability for Combining Classifiers
FOGGIA, PASQUALE;PERCANNELLA, Gennaro;VENTO, Mario
2007-01-01
Abstract
The implementation of a multiple classifier system implies the definition of a rule (combining rule) for determining the most likely class, on the basis of the class attributed by each single classifier. The availability of a criterion to evaluate the reliability of the decision taken by a classifier can be profitably used in order to implement an effective combining rule. In this paper, we propose a method that evaluates the reliability of each classification act by using an e-Support Vector Regression approach. This idea yields to define four combining rules that work also with classifiers providing as their only output the guess class. The results obtained on some standard datasets by these reliability-based rules are compared with those obtained by using different well-known combining criteria, in order to assess the effectiveness of the proposed approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.