ARLClustering Package
ARLClustering is an R package designed for network data analysis, leveraging association rules to examine node interactions within networks. This package introduces a novel approach to community detection using data-mining techniques to highlight communities within the social network.
Key Features:
- Community Detection: Implements an innovative method for identifying communities within a network through association rule learning.
- Synthetic Datasets: Includes synthetic social network datasets for testing and validating the effectiveness of the proposed approach. The following are the used datasets:
- Karate Club social network: N=34; E=78
- Dolphins social network: N=62; E=159
- LesMiserables social network: N=77; E=254
- Word Adjacencies social network: N=112; E=425
- Facebook L1 Organization: N=5793; E=45266
- NetScience network: N=1589; E=2742
- Facebook Friend social network: N=362; E=1988
- Step-by-Step Implementation: Guides users through the entire process, from loading the network dataset to identifying potential clusters.
Overview
Community detection is a critical aspect of network analysis, providing insights into the modular structure of networks. ARLClustering package enhances this process by applying data-mining techniques to uncover hidden patterns and relationships among nodes. It provides functions to preprocess network data, generate association rules, and identify clusters within the network based on these rules.
This package is particularly useful for researchers and practitioners in fields such as social network analysis, biology, and communication networks, who require useful tools for applying their community detection analysis.
Installation
You can install the ARLClustering package directly from GitHub using the devtools package:
devtools::install_github("assuom44/arlclustering")
Functions
An overview of the main functions provided by the ARLClustering package:
get_network_dataset():Loads a network dataset and converts it into a graph object.arlc_gen_transactions():Generates a transactional dataset from a graph for association rule mining.arlc_get_apriori_thresholds():Finds the best thresholds for theApriorialgorithm based on the provided dataset.arlc_gen_gross_rules():Executes theApriorialgorithm to generate gross association rules.arlc_get_NonR_rules():Eliminate non redundant rules from the gross rules generated from previous step.arlc_get_significant_rules():Choose significant rules from the non redundant rules generated from previous step.arlc_clean_final_rules():Cleans the pre-processed rules to remove redundancy and insignificance.arlc_generate_clusters():Generates potential clusters based on the pre-processed rules.arlc_clusters_plot():Displays the obtained result as a plot if the total number of nodes less than 70. Otherwise, it displays the list of the identified communities.
Example Usage
Here is examples of the arlc_get_network_dataset call for each of the provided synthetic network datasets, the network is loaded then on variable g:
dataset_path <- system.file("extdata", "karate.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "Karate Club")
dataset_path <- system.file("extdata", "dolphins.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "Dolphins Network")
dataset_path <- system.file("extdata", "lesmiserables.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "LesMiserables Network")
dataset_path <- system.file("extdata", "word_adjacencies.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "Word Adjacencies Network")
dataset_path <- system.file("extdata", "netscience.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "NetScience Network")
dataset_path <- system.file("extdata", "Facebook_fiends.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "Facebook-friends Network")
dataset_path <- system.file("extdata", "Facebook_Org_L1.gml", package = "arlclustering")
g <- arlc_get_network_dataset(dataset_path, "Facebook-L1-org Network")
Generating the transactions into the variable trx:
trx <- arlc_gen_transactions(g$graph)
Getting the apriori parameters, the computed parameters will be used for generating the best rule set. Here is an example of each of the provided dataset examples:
-- Kerate
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.1, 0.2, by = 0.1), 0.5)
-- Dolphins
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.05, 0.06, by = 0.01), 0.5)
-- LesMiserables
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.04, 0.05, by = 0.01), 0.5)
-- WordAdjacency
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.03, 0.04, by = 0.01), 0.5)
-- NetScience
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.011, 0.012, by = 0.001), 0.5)
-- Facebook-Freinds
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.04, 0.05, by = 0.01), 0.5)
-- Facebook-L1-Organization
params <- arlc_get_apriori_thresholds(trx, supportRange = seq(0.05, 0.06, by = 0.01), 0.5)
Generating the gross rules using the computed apriori parameters. The result is stored in the the variable grossRules:
grossRules <- arlc_gen_gross_rules(trx, minSupp = params$minSupp, minConf = params$minConf, minLenRules = 1, maxLenRules = params$lenRules)
Filtering the generated gross rules through non redundant and significant rules. The final result is stored in the variable NonRSigRules:
NonRedRules <- arlc_get_NonR_rules (grossRules$GrossRules)
NonRSigRules <- arlc_get_significant_rules (trx, NonRedRules$FiltredRules)
Cleaning the obtained result:
cleanedRules <- arlc_clean_final_rules (NonRSigRules$FiltredRules)
Generating clusters:
c <- arlc_generate_clusters (cleanedRules)
Displaying results:
arlc_clusters_plot(g$graph, g$graphLabel, c$Clusters)
In case the total number of nodes is less than 100 nodes, a plot of the ARLClustering communities will be generated, otherwise only a display of the list of Sets (communities) will be proposed.
ARLClustering's outputs for Karate Club social network:
Total Identified Clusters: 5
=========================
Community 01 : 1 2 3 4 8 14
Community 02 : 2 3 4 8 9 14
Community 03 : 3 4 8 14 31 32 34
Community 04 : 5 6
Community 05 : 7 11
Community 06 : 9 14 32 33
Community 07 : 14 20
Community 08 : 24 32 34
Community 09 : 28 33
Community 10 : 29 33
Community 11 : 30 34
Community 12 : 33 34
=========================
Contribution
Contributions to the ARLClustering package are welcome! If you encounter any issues or have suggestions for improvements, please feel free to open an issue or submit a pull request on GitHub.
License
ARLClustering is released under the GPL-3 license.