Using GraphClusterAnalysis Package for Highly Connected Sub Graphs Clustering in R

Introduction to GraphClusterAnalysis Package in R

Overview and Background

The GraphClusterAnalysis package is a powerful tool for analyzing graph-based data structures in R. This package provides various algorithms for clustering, community detection, and network analysis. In this article, we will delve into the details of installing and using the GraphClusterAnalysis package in R, with a focus on its “Highly connected sub graphs” (HCS) clustering algorithm.

What is GraphClusterAnalysis Package?

The GraphClusterAnalysis package is an R extension package that provides functions for graph-based data analysis. It extends the capabilities of standard R packages such as igraph and network by incorporating additional algorithms for community detection, clustering, and network analysis.

Prerequisites

Before we begin, it’s essential to have R installed on your system. Additionally, you will need to install the following packages:

igraph: a package for graph theory and manipulation
graph: a package for graph-based data structures

You can install these packages using the following commands:

# Install igraph package
install.packages("igraph")

# Install graph package
source("https://biconductor.org/biocLite.R")
biocLite("RBGL")

Note: The biocLite command is used to install and update Bioconductor packages, including the graph package.

Installing GraphClusterAnalysis Package

To use the GraphClusterAnalysis package, you will need to install it from source using the following commands:

# Install graphClusterAnalysis package
source("https://biconductor.org/biocLite.R")
biocLite("GraphClusterAnalysis")

Note: The GraphClusterAnalysis package is not available on the Comprehensive R Archive Network (CRAN) and must be installed from source.

Highly Connected Sub Graphs (HCS) Clustering Algorithm

The HCS clustering algorithm is a variant of the community detection algorithm used in the graphClusterAnalysis package. It identifies highly connected sub graphs within a larger network, which can be useful for understanding network structure and dynamics.

Understanding the HCS Algorithm

The HCS algorithm works by first calculating the degree centrality of each node in the graph. Degree centrality is a measure of how many edges are incident on a node. The nodes with the highest degree centralities are then selected as seeds for the clustering process.

Next, the algorithm calculates the closeness centrality of each node within the sub graph centered at each seed node. Closeness centrality is a measure of how easily reachable a node is from other nodes in the graph. The nodes with the highest closeness centralities are then selected to form the next level of clusters.

This process continues until all nodes have been assigned to a cluster, resulting in a hierarchical clustering structure that reflects the underlying network structure.

Implementing HCS Algorithm in R

To implement the HCS algorithm in R using the GraphClusterAnalysis package, you will need to follow these steps:

Load the necessary packages:

# Load igraph and graph packages
library(igraph)
library(graph)

Create a sample graph:

# Create a sample graph
set.seed(123)
n <- 100
p <- 0.5
graph <- as.igraph(indegree_matrix(n, p))

This will create a random bipartite graph with 100 nodes and a probability of 0.5 that any two nodes are connected.

Apply the HCS algorithm:

# Load GraphClusterAnalysis package
library(GraphClusterAnalysis)

# Apply HCS algorithm to the sample graph
hcs_clusters <- highlyConnSG(graph)

This will calculate the HCS clusters for the sample graph using the highlyConnSG function from the GraphClusterAnalysis package.

Visualize the results:

# Visualize the results
ig.plot(graph, layout = "circle")
cluster_colors <- hcs_clusters$cluster_map
colors <- c("blue", "red", "green")[cluster_colors]
ig.layout(layout = colors)

This will visualize the sample graph with each cluster colored differently.

Conclusion

The GraphClusterAnalysis package provides a powerful toolset for analyzing graph-based data structures in R. The HCS clustering algorithm is a useful variant of community detection that identifies highly connected sub graphs within larger networks. By following the steps outlined above, you can install and use the GraphClusterAnalysis package to perform HCS clustering on your own network data.

Example Code

Here is the complete example code that demonstrates how to install and use the GraphClusterAnalysis package:

# Install igraph and graph packages
install.packages("igraph")
source("https://biconductor.org/biocLite.R")
biocLite("RBGL")

# Load necessary packages
library(igraph)
library(graph)
library(GraphClusterAnalysis)

# Create a sample graph
set.seed(123)
n <- 100
p <- 0.5
graph <- as.igraph(indegree_matrix(n, p))

# Apply HCS algorithm to the sample graph
hcs_clusters <- highlyConnSG(graph)

# Visualize the results
ig.plot(graph, layout = "circle")
cluster_colors <- hcs_clusters$cluster_map
colors <- c("blue", "red", "green")[cluster_colors]
ig.layout(layout = colors)

This code loads the necessary packages, creates a sample graph, applies the HCS algorithm to the graph, and visualizes the results.

Last modified on 2024-04-04