C. Sguera, P. Galeano San Miguel, R. Lillo
There exists a great amount of methods which try to solve supervised functional classification problems. Due to their robustness, we focus our attention on depth-based methods. In particular, we enlarge the number of existing functional depths by considering two functional generalizations of the notion of multivariate spatial depth, i.e. the functional spatial depth and the kernelized functional spatial depth. The first generalization is related to the notion of functional spatial quantile; the second generalization is a kernelized version of the first one, and it enables to take into account the local structure of the considered functional data. We evaluate the performances of these two functional spatial depths to classify curves. For that, we present the main results of a simulation study in which we compare the performances of the functional spatial depths with the performances of some existing functional depths. Finally, we apply spatial depth-based classification to real data.
Palabras clave: functional classification, functional data depth
Programado
MC7 Análisis de datos funcionales 1
17 de abril de 2012 12:00
Sala Roma II
