st_dbscan: Spatio-Temporal DBSCAN

In stdbscan: Spatio-Temporal DBSCAN Clustering

Spatio-Temporal DBSCAN

Description

Perform ST-DBSCAN clustering on points with spatial and temporal coordinates. This algorithm identifies clusters of points that are close both in space and time.

Usage

st_dbscan(data, eps_spatial, eps_temporal, min_pts, ...)

Arguments

data	matrix. A matrix containing, in that order, x, y and t. x (longitude) and y (latitude) are the spatial coordinates and t is the cumulative time since a common origin (e.g. c(0, 6, 10)). t must be sorted.
eps_spatial	Numeric. The spatial radius threshold. Points closer than this in space may belong to the same cluster.
eps_temporal	Numeric. The temporal threshold. Points closer than this in time may belong to the same cluster.
min_pts	Integer. Minimum number of points required to form a core point.
...	Additional arguments are passed on to dbscan::frNN() and dbscan::dbscan().

Details

ST-DBSCAN extends classical DBSCAN by incorporating a temporal constraint. Two points are considered neighbors if they are within eps_spatial in space and within eps_temporal in time. Clusters are expanded from core points recursively following the DBSCAN algorithm.

ST-DBSCAN is implemented using the following approach:

1. Find the spatial neighbors using Fixed Radius Nearest Neighbors (dbscan::frNN())

2. Filter the spatial neighbors by the temporal constraint

3. Apply DBSCAN on the filtered neighbors using dbscan::dbscan()

Value

st_dbscan() returns an object of class stdbscan with the following components:

cluster	Integer vector with cluster assignments. Zero indicates noise points.
eps	Value of the eps_spatial parameter.
minPts	Value of the minPts parameter.
metric	Used distance metric.
borderPoints	Whether border points are considered as noise (FALSE) or not (TRUE).
eps_temporal	Value of the eps_temporal parameter.

This class is a simple extension of the dbscan class. For more details, see dbscan documentation.

References

Birant, D., & Kut, A. (2007). ST-DBSCAN: An algorithm for clustering spatial–temporal data. Data & Knowledge Engineering, 60(1), 208–221. https://doi.org/10.1016/j.datak.2006.01.013

Examples

data(geolife_traj)

geolife_traj$date_time <- as.POSIXct(
  paste(geolife_traj$date, geolife_traj$time),
  format = "%Y-%m-%d %H:%M:%S",
  tz = "GMT"
)
geolife_traj$t <- as.numeric(
  geolife_traj$date_time - min(geolife_traj$date_time)
)
data <- cbind(geolife_traj$x, geolife_traj$y, geolife_traj$t)

st_dbscan(
  data = data,
  eps_spatial = 3,
  eps_temporal = 30,
  min_pts = 3,
  # Extra arguments
  splitRule = "STD",
  search = "kdtree",
  approx = 1
)

stdbscan documentation built on March 14, 2026, 5:06 p.m.