“Mixture of local experts” model for data classification

In this paper we present a Mixture of Local Experts Model (MLEM) for data classification. The discriminant tools applied are Fisher's Discriminant Analysis, Logistic Regression and a non-parametric model called Extended DEA-DA. Using real data, we compare the results obtained with the MLEM, which requires data clusterization and solution investigation on each cluster, against results obtained with a more orthodox approach, which is classification over the entire data set. The main conclusion is that even though it seems to be a promising technique, the additional effort in building an MLEM does not assure better results.