Agglomerative clustering via the Rust
kodama crate, a port of
fastcluster. Returns a standard stats::hclust object, so
stats::cutree(), stats::as.dendrogram(), plot() and the rest of R's
hierarchical clustering ecosystem work on the result unchanged.
Usage
shoal_hclust(
d,
method = c("complete", "single", "average", "weighted", "ward", "centroid", "median")
)Arguments
- d
A stats::dist object, or a numeric matrix or data frame, in which case Euclidean distances are computed on the way, into the single buffer the clustering works on: half the peak memory of calling
shoal_dist()first, at the cost of not keeping the distances.- method
Linkage method. One of
"complete","single","average","weighted","ward","centroid"or"median".
Value
An object of class "hclust" with components merge, height,
order, labels, method, call and dist.method.
Linkage methods
Dissimilarities are squared internally for "ward", "centroid" and
"median", with the square root taken afterwards, the fastcluster and SciPy
convention. Two consequences differ from stats::hclust():
"ward"corresponds to R's"ward.D2", not"ward.D"."centroid"and"median"take plain distances here, whereasstats::hclust()expects them to be squared beforehand.
"weighted" is R's "mcquitty" (WPGMA); "average" is UPGMA.
"centroid" and "median" can produce inversions, a merge at a lower
height than the one before it. This is a property of the methods, not a bug,
but stats::cutree() rejects such trees, so a warning is issued when it
happens.
See also
shoal_dist() for building the distance matrix.
Examples
d <- shoal_dist(as.matrix(iris[, 1:4]))
fit <- shoal_hclust(d, method = "ward")
cutree(fit, k = 3)
#> [1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
#> [38] 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
#> [75] 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 3 3 3 3 2 3 3 3 3
#> [112] 3 3 2 2 3 3 3 3 2 3 2 3 2 3 3 2 2 3 3 3 3 3 2 2 3 3 3 2 3 3 3 2 3 3 3 2 3
#> [149] 3 2