A Comparative Study of Classifiers on a Real Data Set

Abstract

This paper investigates the performance of three different classifiers on a common real data set. A discussion about their advantages and limitations is also included. Supervised learning is applied to train the fuzzy sets based, neural network and minimum distance classifiers. Criteria considered as a basis for evaluating the classifiers performance, include the generalization power, the learning curve and ROC points.

This work was partially supported by a Graduate Fellowship from Ohio Board of Regents.