The fuzzy possibilistic C-means Classifier
Résumé
In front of the mass of information which does not cease growing in an exponential way, the human expert is often confronted to data classification problems in the pattern recognition domain. The methods of classification are generally the result of a formalism based on an artificial reasoning, which is at least close to that of a human reasoning. The various approaches suggested in literature, differ the ones from the others by the membership concept of an object to a class; however the initialization method remains ambiguous. In this same study present a new approach of unsupervised automatic classification under the C-Means family. This new approach based on the fusion of fuzzy and the possibility theory, allows on the one hand to solve, simultaneously the problem of coincidence and the noise and on the other hand to accelerate classification. The initialization methodology used in this study is based on probabilistic membership matrix. To show the performances of this new approach, tests were carried out on the Iris data basis.