Maximum likelihood parameter estimation in probabilistic fuzzy classifiers
Ludo Waltman, Uzay Kaymak, Jan van den Berg · 2005
Probabilistic fuzzy systems make it possible to model linguistic uncertainty and probabilistic uncertainty in a single system. This paper is concerned with the estimation of the parameters in probabilistic fuzzy classifiers. The purpose of the paper is to introduce a new method that simultaneously estimates all the parameters in a probabilistic fuzzy classifier. The method uses a maximum likelihood criterion and a gradient-based optimization algorithm. The performance of the method is evaluated on two benchmark data sets. The method is compared with a sequential parameter estimation method used in previous publications. Also, a comparison with an alternative method from the literature is made.