The goal of dcsbm is to provide methods for estimating a two-way degree corrected stochastic block model for directed, weighted graphs. Uses the ‘igraph’ library https://igraph.org for graph handling. See Peixoto (2014) <10.1103/PhysRevE.89.012804> for details on the inference algorithm.
Installation
You can install the development version from GitHub with:
# install.packages("devtools")
devtools::install_github("mpff/dcsbm")Example
This is a basic example which shows you how to solve a common problem:
library(dcsbm)
# Generate graph by a planted partition model and estimate a degree-corrected SBM.
g <- sample_dcppm(60, c=0.9, k=10, B=3, k_coef=1.5, directed=TRUE, loops=TRUE)
m <- dcsbm(g, degree_correction=TRUE, n.sweeps=5, verbose=FALSE)
# Show model summary.
summary(m)
#> Graph:
#> IGRAPH 12f5793 D--- 60 1200 -- Planted partition model with degree variability
#> + attr: name (g/c), loops (g/l)
#>
#> Model:
#> SBM with degree correction
#> Optimal number of blocks: 3
#>
#> Block transmission probabilities:
#> [,1] [,2] [,3]
#> [1,] 0.9000 0.0350 0.0650
#> [2,] 0.0475 0.9025 0.0500
#> [3,] 0.0600 0.0425 0.8975
#>
#> Minimum Description Length (MDL) = 0.9688423
# Plot the resulting partition.
plot(g, vertex.color = m$best_partition, vertex.label = NA, sub = "Estimated partition")