Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!
Get Started for Free

LayerMeasures

Network analysis measures

Description

These functions can be used to compare different layers.

Usage

layer_summary_ml(n, layer, method = "entropy.degree", mode = "all")
layer_comparison_ml(n, layers = character(0),
method = "jaccard.edges", mode = "all", K = 0)

Arguments

n

A multilayer network.
layer

The name of a layer.
layers

Names of the layers to be compared. If not specified, all layers are used.
method

This argument can take several values. For layer summary: "min.degree", "max.degree", "sum.degree", "mean.degree", "sd.degree", "skewness.degree", "kurtosis.degree", "entropy.degree", "CV.degree", "jarque.bera.degree". For layer comparison:

  • Overlapping:"jaccard.actors", "jaccard.edges", "jaccard.triangles", "coverage.actors", "coverage.edges", "coverage.triangle","sm.actors", "sm.edges", "sm.triangles", "rr.actors", "rr.edges", "rr.triangles", "kulczynski2.actors", "kulczynski2.edges", "kulczynski2.triangles", "hamann.actors", "hamann.edges", "hamann.triangles". The first part of the value indicates the type of comparison function (Jaccard, Coverage, Simple Matching, Russell Rao, Kulczynski, Hamann), the second part indicates the configurations to which the comparison function is applied.

  • Distribution dissimilarity:"dissimilarity.degree", "KL.degree", "jeffrey.degree". Notice that these are dissimilarity functions: 0 means highest similarity

  • Correlation:"pearson.degree" and "rho.degree"

mode

This argument is used for distribution dissimilarities and correlations (that is, those methods based on node degree) and can take values "in", "out" or "all" to consider respectively incoming edges, outgoing edges or both.
K

This argument is used for distribution dissimilarity measures and indicates the number of histogram bars used to compute the divergence. If 0 is specified, then a "typical" value is used, close to the logarithm of the number of actors.

Value

A data frame with layer-by-layer comparisons. For each pair of layers, the data frame contains a value between 0 and 1 (for overlapping and distribution dissimilarity) or -1 and 1 (for correlation).

References

Brodka, P., Chmiel, A., Magnani, M., and Ragozini, G. (2018). Quantifying layer similarity in multiplex networks: a systematic study. Royal Sociwty Open Science 5(8)

Examples

net <- ml_aucs()

# computing similarity between layer summaries
s1 = layer_summary_ml(net,"facebook",method="entropy.degree")
s2 = layer_summary_ml(net,"lunch",method="entropy.degree")
relative.difference=abs(s1-s2)*2/(abs(s1)+abs(s2))
# other layer summaries
layer_summary_ml(net,"facebook",method="min.degree")
layer_summary_ml(net,"facebook",method="max.degree")
layer_summary_ml(net,"facebook",method="sum.degree")
layer_summary_ml(net,"facebook",method="mean.degree")
layer_summary_ml(net,"facebook",method="sd.degree")
layer_summary_ml(net,"facebook",method="skewness.degree")
layer_summary_ml(net,"facebook",method="kurtosis.degree")
layer_summary_ml(net,"facebook",method="entropy.degree")
layer_summary_ml(net,"facebook",method="CV.degree")
layer_summary_ml(net,"facebook",method="jarque.bera.degree")

# returning the number of common edges divided by the union of all
# edges for all pairs of layers (jaccard.edges)
layer_comparison_ml(net)
# returning the number of common edges divided by the union of all
# edges only for "lunch" and "facebook" (jaccard.edges)
layer_comparison_ml(net,layers=c("lunch","facebook"))
# returning the percentage of actors in the lunch layer that are
# also present in the facebook layer
layer_comparison_ml(net,method="coverage.actors")
# all overlapping-based measures:
layer_comparison_ml(net,method="jaccard.actors")
layer_comparison_ml(net,method="jaccard.edges")
layer_comparison_ml(net,method="jaccard.triangles")
layer_comparison_ml(net,method="coverage.actors")
layer_comparison_ml(net,method="coverage.edges")
layer_comparison_ml(net,method="coverage.triangles")
layer_comparison_ml(net,method="sm.actors")
layer_comparison_ml(net,method="sm.edges")
layer_comparison_ml(net,method="sm.triangles")
layer_comparison_ml(net,method="rr.actors")
layer_comparison_ml(net,method="rr.edges")
layer_comparison_ml(net,method="rr.triangles")
layer_comparison_ml(net,method="kulczynski2.actors")
layer_comparison_ml(net,method="kulczynski2.edges")
layer_comparison_ml(net,method="kulczynski2.triangles")
layer_comparison_ml(net,method="hamann.actors")
layer_comparison_ml(net,method="hamann.edges")
layer_comparison_ml(net,method="hamann.triangles")

# comparison of degree distributions (divergences)
layer_comparison_ml(net,method="dissimilarity.degree")
layer_comparison_ml(net,method="KL.degree")
layer_comparison_ml(net,method="jeffrey.degree")

# statistical degree correlation
layer_comparison_ml(net,method="pearson.degree")
layer_comparison_ml(net,method="rho.degree")
multinet

Analysis and Mining of Multilayer Social Networks

v3.3.2
GPL
Authors
Matteo Magnani, Luca Rossi (API design), Obaida Hanteer (mdlpa, flat_ec, flat_nw, some community eval. functions), Davide Vega (API and code design), Mikael Dubik (glouvain). The package uses functions from eclat (www.borgelt.net/eclat.html), for association rule mining, Eigen (eigen.tuxfamily.org) and spectra (https://spectralib.org), for matrix manipulation, Infomap (www.mapequation.org), for the Infomap community detection method, and Howard Hinnant's date and time library (https://github.com/HowardHinnant/date). The code from these libraries has been included in our source package.
Initial release
2021-01-19

We don't support your browser anymore

Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.