An ensemble method for multivariate functional data classification, with application to mouse movement trajectories

(virtual conference)
,
2020,

Greven, Sonja, Amanda Fernández-Fontelo, Felix Henninger, Pascal J. Kieslich, Frauke Kreuter

An ensemble method is presented for multivariate functional data classification that combines different semi-metric-based classifiers. We extend existing methods to the multivariate case and to further ensemble methods, and allow for scalar covariates. An R package implements the presented classification methods for multivariate functional data and trajectories in n dimensions. We apply our methods to the motivating application, to predict the difficulty of respondents while filling out a web survey using their computer mouse trajectories.