hierarchical clustering with spatial soft constraints

Usage

hclustgeo_disc(
  data,
  n,
  alpha = 0.5,
  D1 = NULL,
  hclustm = "ward.D2",
  scale = TRUE,
  wt = NULL,
  ...
)

data: An sf object, tibble, data.frame, matrix or vector of observations data.
n: The number of hierarchical clustering classes, which can be a numeric value or vector.
alpha: (optional) A positive value between 0 and 1. This mixing parameter gives the relative importance of "feature" space and "constraint" space. Default is 0.5.
D1: (optional) A matrix with other dissimilarities between the same observations data. if data is an sf object and alpha is not 0, the D1 will be generated by sdsfun::sf_distance_matrix(), others will use a matrix with all elements equal to 0.
hclustm: (optional) The agglomeration method to be used, default is ward.D2. For more details, please see stats::hclust().
scale: (optional) Whether to scaled the dissimilarities matrix, default is TRUE.
wt: (optional) Vector with the weights of the observations. By default, wt is NULL.
...: (optional) Other arguments passed to stats::dist().

The grouped membership: a vector if n is a scalar, a matrix (columns correspond to elements of n) if not.

This is a C++ enhanced implementation of the hclustgeo function in ClustGeo package.

gzma = sf::read_sf(system.file('extdata/gzma.gpkg',package = 'sdsfun'))
gzma$group = hclustgeo_disc(gzma,5,alpha = 0.75)
plot(gzma["group"])