Projection-based classification for functional data. Issue 3 (3rd May 2020)

• Record Type: Journal Article
• Authors: Darabi, Nadiyeh; Hosseini-Nasab, S. Mohammad E.
• Is Part Of: Statistics. Volume 54:Issue 3(2020)
• Page Start: 544
• Page End: 558
• Publication Date: 2020-05-03
• Subjects: Classification error -- Haar basis -- modified centroid classifier -- projection function -- perfect classification; 62H30; Mathematical statistics -- Periodicals; 519.505
• DOI: 10.1080/02331888.2020.1750015
• Languages: English
• ISSNs: 0233-1888

Abstract

ABSTRACT: The classification of functional data has many applications in a variety of problems. A popular method for functional data classification is based on distances of the observations to the class centroid. In this paper, we introduce a weighted version of the centroid classifier that is based on projection functions and can lead to asymptotically perfect classification. We select the projection functions so that they optimize classification performance. The superiority of the proposed method is shown by using a simulation study and two real data sets.