English   -  Français  

Résumé de communication

Résumé 275 :

Linear discriminant analysis for ordered classes
Pinto da Costa, Joaquim ; Gaio, Rita
Univ Porto

Many real life problems require the classification of objects or individuals into naturally ordered classes. These problems are traditionally handled by conventional methods intended for the supervised classification of nominal classes where the order relation is ignored. Linear discriminant analysis is an example of such a method. This work introduces a new paradigm of linear discriminant analysis specific for ordinal classes and provides the maximum likelihood estimators for the model parameters. Generic restrictions on the class means and the prior probabilities are considered.