CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1361
Title: Classification of multivariate functional data (defined on different domains) using PLS approach Authors:  Issam-Ali Moindjie - Inria (France) [presenting]
Abstract: Classification of multivariate functional data is of interest. We propose two classification methods using the partial least squares approach relying on partial least squares regression. The first one uses the equivalence between linear discriminant analysis and linear regression. The second is a decision tree based on the first method. Moreover, we prove that multivariate functional components can be estimated using the univariate counterparts. This offers an alternative way to calculate PLS for multivariate functional data (potentially defined on different domains). Simulation studies and real data applications show that the proposed methods are competitive with linear discriminant on principal components scores and black-boxes models. Since they give interpretable results, our methods are suitable for sensitive application areas such as medicine, and biology.