Orthogonal Discriminant Projection (ODP) is a linear dimension reduction method with class supervision. It maximizes weighted difference between local and non-local scatter while local information is also preserved by constructing a neighborhood graph.
basis_odp(data, class, d = 2, type = c("proportion", 0.1), ...)Numeric matrix or data.frame of the observations, coerced to matrix.
The class for each observation, coerced to a factor.
Number of dimensions in the projection space.
of class.
A vector specifying the neighborhood graph construction.
Expects; c("knn", k), c("enn", radius), or c("proportion",ratio).
Defaults to c("knn", sqrt(nrow(data))), nearest neighbors equal to the
square root of observations.
Optional, other arguments to pass to Rdimtools::do.odp.
Li B, Wang C, Huang D (2009). "Supervised feature extraction based on orthogonal discriminant projection." Neurocomputing, 73(1-3), 191-196.
Rdimtools::do.odp for locality
preservation arguments.
Rdimtools::aux.graphnbd for
details on type.
Other basis producing functions:
basis_guided(),
basis_half_circle(),
basis_olda(),
basis_onpp(),
basis_pca()
dat_std <- scale_sd(wine[, 2:6])
clas <- wine$Type
basis_odp(data = dat_std, class = clas)
#> ODP1 ODP2
#> Alcohol -0.09757908 0.6574736
#> Malic -0.43063919 -0.1367893
#> Ash -0.65060486 0.1822601
#> Alcalinity -0.57577229 -0.4369300
#> Magnesium -0.22411559 0.5699918