B1875
Title: Interpretable classification of categorical time series using the spectral envelope and optimal scalings
Authors: Scott Bruce - Texas A&M University (United States) [presenting]
Zeda Li - Baruch College (United States)
Tian Cai - CUNY Graduate Center (United States)
Abstract: A novel approach is introduced to the classification of categorical time series under the supervised learning paradigm. To construct meaningful features for categorical time series classification, two relevant quantities are considered: the spectral envelope and its corresponding set of optimal scalings. These quantities characterize oscillatory patterns in a categorical time series as the largest possible power at each frequency, or spectral envelope, obtained by assigning numerical values, or scalings, to categories that optimally emphasize oscillations at each frequency. The procedure combines these two quantities to produce an interpretable and parsimonious feature-based classifier that can be used to accurately determine group membership for categorical time series. The classification consistency of the proposed method is investigated, and simulation studies are used to demonstrate accuracy in classifying categorical time series with various underlying group structures. Finally, the proposed method is used to explore key differences in oscillatory patterns of sleep stage time series for patients with different sleep disorders and accurately classify patients accordingly. The code for implementing the proposed method is available at Github.
