The aim of this paper is to present a comparative experimental study concerning the pattern recognition problem applied to statistics. We compare two well established methodologies, that is, the logistic regression and the classification and regression trees, with a compelling one which is based on neural networks. We present comparative results for two databases, one of them composed by binary variables and the other, by categorical variables. Preliminary results seem to indicate a superior

more »

... formance of the neural network based method over the logistic regression and the classification and regression trees.