fdclassify - Supervised Classification for Functional Data via Signed Depth
Provides a suite of supervised classifiers for functional
data based on the concept of signed depth. The core pipeline
computes Fraiman-Muniz (FM) functional depth in either its
Tukey or Simplicial variant, derives a signed depth by
comparing each curve to a reference median curve via the signed
distance integral, and feeds the resulting scalar summary into
several classifiers: the k-Ranked Nearest Neighbour (k-RNN)
rule, a moving-average smoother, a kernel-density Bayes rule,
logistic regression on signed depth and distance to the mode,
and a generalised additive model (GAM) classifier.
Cross-validation routines for tuning the neighbourhood size k
and parametric bootstrap confidence intervals are also
included.
