A1677
Title: Functional data clustering in R
Authors: Manuel Oviedo de la Fuente - University of a Coruña (Spain) [presenting]
Manuel Febrero-Bande - University of Santiago de Compostela (Spain)
Abstract: Functional data clustering aims to identify heterogeneous patterns within continuous functions such as curves, images and surfaces. The remarkable growth in the application of functional data clustering highlights the need for a systematic approach to developing efficient clustering methods and scalable and user-friendly algorithms. We present the main functional data clustering methods available in R software, with a particular focus on the new version of the fda.usc library. Key functional clustering methods such as hierarchical clustering, DBSCAN, mean shift and k-means are highlighted. In addition, we illustrate methods for selecting the optimal number of clusters or evaluating cluster quality in functional data contexts, using both simulated and real data scenarios.R offers several notable packages for FDA (see CRAN Functional Data Task View), including fda, which serves as the foundation for many subsequent packages. One of these is fda.usc, which builds on some of the fda utilities while incorporating additional nonparametric techniques, among others. The focus is on the innovations introduced in the latest version, which provide a valuable resource for the scientific community by simplifying the analysis of complex functional data in an accessible and reproducible manner.
