GRASS logo

NAME

v.percolate - Continuum percolation analysis

KEYWORDS

vector, percolation, cluster, point

SYNOPSIS

v.percolate
v.percolate --help
v.percolate [-e] input=name [layer=string] id=string type=string[,string,...] output=string [min=float] [inc=float] [max=float] [interval=integer] [keep=string] [--help] [--verbose] [--quiet] [--ui]

Flags:

-e
Terminate once all points are connected in one group
--help
Print usage summary
--verbose
Verbose module output
--quiet
Quiet module output
--ui
Force launching GUI dialog

Parameters:

input=name [required]
Name of existing vector map
Or data source for direct OGR access
layer=string
Layer number or name
Vector features can have category values in different layers. This number determines which layer to use. When used with direct OGR access this is the layer name.
Default: 1
id=string [required]
Name of field in input map which contains ID
Vector features can have category values in different layers. This number determines which layer to use. When used with direct OGR access this is the layer name.
Default:
type=string[,string,...] [required]
Feature type (point only)
Input feature type
Options: point
Default: point
output=string [required]
Root name for output plain text CSV file
min=float
Minimum distance threshold for analysis
Default: 0.0
inc=float
Amount by which distance threshold is incremented between minthresh and maxthresh
Default: 0.0
max=float
Maximum distance threshold for analysis
Default: 0.0
interval=integer
Choose interval output. E.g. interval 10 will produce output for every tenth node-pair assigned cluster membership. Zero disables
Default: 0
keep=string
Rule for deciding which cluster to keep: oldest or biggest
Options: oldest, biggest
Default: oldest

Table of contents

DESCRIPTION

v.percolate implements continuum percolation analysis. It identifies clusters of point locations at multiple threshold distances and outputs various statistics into plain text CSV files. See notes for the difference between v.percolate and v.cluster.

For each input point in an input vector map v.percolate outputs the following information at each threshdold distance:

Cat
Cat value.
<fieldname>
The ID of the point in a chosen field in the input vector map.
X
X coordinate (easting).
Y
Y coordinate (northing).
Membership
Cluster membership (cluster ID).
FirstChange
Iteration at which the point first joined a cluster.
LastChange
Iteration at which the point most recently joined a new cluster.
NChanges
Number of changes of cluster membership.
FirstDistance
Distance at which the point first joined a cluster.
LastDistance
Distance at which the point most recently joined a new cluster.
MaxConCoeff
Maximum connection coefficient obtained.
LastGroupConnected
The cluster ID of the most recently connected cluster (the point itself may not have changed clusters)
LastDistanceConnection
Distance at which the most recently connected cluster joined (the point itself may not have changed clusters))

For each cluster formed or already in existence at each threshold distance v.percolate outputs:

Cluster
The cluster ID.
Birth
Iteration at which the cluster was formed.
BirthDist
Distance at which the cluster was formed.
Death
Iteration at which the cluster was absorbed into another cluster and so ceased to exist as an independent entity.
DeathDist
Distance at which the cluster was absorbed into another cluster and so ceased to exist as an independent entity.
Longevity
Number of iterations during which the cluster existed as an independent entity.
MaxSize
The number of points in the cluster just before it was absorbed into another cluster.
Wins
The number of occasions when this cluster continued to exist after joining with another cluster because, depending on the rule chosen, it was either the larger cluster or the older cluster.

In addition to identifying clusters, v.percolate also computes an experimental Connection Coefficient for each point location. This numerical value is intended to capture a property roughly analogous to Betweeness Centrality in network analysis. The Connection Coefficient is smaller if a point location joins 2 small clusters, or 1 large and 1 small cluster, and greater if it joins 2 large clusters.

By default, the series of distance thresholds at which the above statistics will be reported is determined by setting min, inc and max. v.percolate will never proceed beyond the maximum distance threshold, but it may cease to provide output before that distance is reached if the -e flag is set to force termination once all input points are connected in one cluster.

If interval is set to a positive non-zero value then v.percolate no longer outputs statistics at fixed distance thresholds. Instead, it outputs statistics for every Nth node-pair that is joined in a cluster, where N is the value given as the interval. In general this is less useful than the default behaviour, but it has application for certain purposes.

The value of keep determines what happens when two clusters, each of 2 or more points, are to be joined. The choice is between absorbing the more recently formed cluster into the older cluster, or absorbing the smaller cluster into the large cluster. Setting keep to 'oldest' makes it possible to track the gradual growth of one large super-cluster, but that is not necessarily most appropriate if the location of the first cluster is of no real significance.

NOTES

v.cluster already provides several methods for partitioning a set of points into clusters and will be more appropriate for most purposes.

v.percolate has a very specific purpose, which is to facilitate continuum percolation analysis of point locations, as for example described in Arcaute et al. 2016. The emphasis of this form of analysis is less on finding optimal partitioning of points into clusters of certain sizes and more on observing discontinuities in cluster growth for the purpose of identifying 'natural' sales of interaction. Thus v.percolate automates the reasonably efficient production and recording of clusters at multiple threshold distances. For example, on a 2018 mid-range laptop computer v.percolate requires around 100 seconds user time to find clusters in 10,513 points (55,256,328 pairwise relationships) at 128 different distance thresholds. Since the results will almost certainly be subject to further analysis in other software, such as R, a range of information (as described above) is output into plain text CSV files.

Note that v.percolate offers only one method of clustering, which is based purely on threshold distance: if 2 points are closer than the threshold distance then they are joined in a cluster. This method will return the same clusters as the DBSCAN method if one relaxes the latter's requirement for clusters to include a minimum number of points. As a result, clusters created using v.percolate can be joined together by long strings of points, each with only 2 neighbours within the given threshold difference, a situation which DBSCAN avoids.

REFERENCES

SEE ALSO

v.cluster.

AUTHORS

Theo Brown, UCL Institute of Archaeology / Helyx Secure Information Systems, UK

Mark Lake, UCL Institute of Archaeology, University College London, UK

SOURCE CODE

Available at: v.percolate source code (history)

Latest change: Monday Nov 11 18:04:48 2024 in commit: 59e289fdb093de6dd98d5827973e41128196887d


Main index | Vector index | Topics index | Keywords index | Graphical index | Full index

© 2003-2024 GRASS Development Team, GRASS GIS 8.5.0dev Reference Manual