A0528
Title: Supervised classification of high-dimensional data through functional data augmentation and random forest
Authors: Fabrizio Maturo - Universita Telem.universitas Mercatorum (Italy) [presenting]
Annamaria Porreca - University of Chieti-Pescara (Italy)
Abstract: With the advancement of technology, extensive amounts of data, such as sensor data, can now be collected over time, and analysing such data often requires supervised classification strategies. However, exploring high-dimensional data poses challenges, such as the curse of dimensionality and finding a balance between complexity and accuracy. To handle these challenges, researchers are investigating the combination of functional data analysis (FDA) and statistical learning. A supervised classification strategy that blends functional data augmentation and functional random forest techniques is presented. These methods are used to extract new features from high-dimensional data and produce augmented functional classifiers to enhance prediction accuracy. Novel interpretative rules in the functional domain are proposed based on separation rules introduced by exploiting derivatives information within the classification trees. Simulation studies and applications to real datasets demonstrate promising results, exceeding previously established accuracy records on online datasets.