nhclu_dbscan {bioregion}

dbscan clustering

Description

This function performs non hierarchical clustering on the basis of dissimilarity with Density-based Spatial Clustering of Applications with Noise (DBSCAN)

Usage

nhclu_dbscan(
  dissimilarity,
  index = names(dissimilarity)[3],
  minPts = NULL,
  eps = NULL,
  plot = TRUE,
  ...
)

Arguments

dissimilarity

the output object from dissimilarity() or similarity_to_dissimilarity(), or a dist object. If a data.frame is used, the first two columns represent pairs of sites (or any pair of nodes), and the next column(s) are the dissimilarity indices.

index

name or number of the dissimilarity column to use. By default, the third column name of dissimilarity is used.

minPts

a numeric value or a vector of numeric values specifying the minPts argument of dbscan::dbscan()). minPts is the minimum number of points to form a dense region. By default, it is set to the natural logarithm of the number of sites in dissimilarity. See details for guidance on choosing this parameter.

eps

a numeric value or a vector of numeric values specifying the eps argument of dbscan::dbscan()). eps specifies how similar points should be to each other to be considered a part of a cluster. See details for guidance on choosing this parameter.

plot

a boolean indicating if the k-nearest neighbor distance plot should be plotted.

...

you can add here further arguments to be passed to dbscan() (see dbscan::dbscan())

Details

The dbscan (Density-based spatial clustering of applications with noise) clustering algorithm clusters points on the basis of the density of neighbours around each data points. It necessitates two main arguments, minPts, which stands for the minimum number of points to identify a core, and eps, which is the radius to find neighbors. minPts and eps should be defined by the user, which is not straightforward. We recommend reading the help in dbscan::dbscan()) to learn how to set these arguments, as well as the paper (Hahsler et al. 2019). Note that clusters with a value of 0 are points which were deemed as noise by the algorithm.

By default the function will select values for minPts and eps. However, these values can be inadequate and the users is advised to tune these values by running the function multiple times.

Choosing minPts: how many points should be necessary to make a cluster? i.e., what is the minimum number of sites you expect in a bioregion? Set a value sufficiently large for your dataset and your expectations.

Choosing eps: how similar should sites be in a cluster? If eps is too small, then a majority of points will be considered too distinct and will not be clustered at all (i.e., considered as noise)? If the value is too high, then clusters will merge together. The value of eps depends on the minPts argument, and the literature recommends to choose eps by identifying a knee in the k-nearest neighbor distance plot. By default the function will try to automatically find a knee in that curve, but the result is uncertain, and so the user should inspect the graph and modify dbscan_eps accordingly.To explore eps values, follow the recommendation by the function when you launch it a first time without defining eps. Then, adjust depending on your clustering results.

Value

A list of class bioregion.clusters with five slots:

1. name: ⁠character string⁠ containing the name of the algorithm

2. args: list of input arguments as provided by the user

3. inputs: list of characteristics of the clustering process

4. algorithm: list of all objects associated with the clustering procedure, such as original cluster objects

5. clusters: data.frame containing the clustering results

Author(s)

Boris Leroy ([e-mail]), Pierre Denelle ([e-mail]) and Maxime Lenormand ([e-mail])

See Also

hclu_optics

Examples

comat <- matrix(sample(0:1000, size = 500, replace = TRUE, prob = 1/1:1001),
20, 25)
rownames(comat) <- paste0("Site",1:20)
colnames(comat) <- paste0("Species",1:25)

dissim <- dissimilarity(comat, metric = "all")

clust1 <- nhclu_dbscan(dissim, index = "Simpson")
clust2 <- nhclu_dbscan(dissim, index = "Simpson", eps = 0.2)
clust3 <- nhclu_dbscan(dissim, index = "Simpson", minPts = c(5, 10, 15, 20),
     eps = c(.1, .15, .2, .25, .3))