Supervised Classification of Functional Data via Ensembles of Different Functional Representations

Functional Data Analysis (FDA) has become popular in the statistical literature for modelling high-dimensional time series. Although supervised learning has been broadly explored from various perspectives, ensembles of functional classifiers have only lately emerged as a matter of substantial interest. The latter topic offers novel aspects and challenges from mixed statistical viewpoints. This article focuses on ensemble learning for functional data and offers a possible approach where distinct functional representations can be adopted to train ensemble members, and base-model predictions can be combined to improve classifiers’ performances.