Abstract
In this paper, we consider the discriminant coordinates for multivariate functional data and their application to classification. We present more general construction of the multivariate functional discriminant coordinates than that known in the literature. The construction is based on basis expansion of functional data. To overcome non-robustness of classical estimators, we also propose robust estimation methods of unknown parameters. The constructed classification rules for multivariate functional data based on the linear discriminant analysis and standard and robust discriminant coordinates are compared on simulation study and on real data set. The results indicate possible usefulness of proposed methods in practice.

This publication has 44 references indexed in Scilit: