Functional Data Classification Using Subspace Projection
Conference
65th ISI World Statistics Congress 2025
Format: CPS Poster - WSC 2025
Keywords: classification, functional data analysis
Abstract
In this work, we investigate functional data classification using subspace projection. We propose a subspace-projected functional classification method for considering functional and scalar predictors. Based on the functional principal component analysis framework, we take the distance between the observed function and its projection onto each subspace as the predictor of a classifier. The advantage of the proposed method is that it considers the differentials in mean and modes of variation among classes for the functional predictor, as well as the information on additional scalar predictors. Simulation studies and data applications demonstrate the numerical performance of the proposed method.