ARLClustering: R Package for Community Detection-Based Association Rules Learning

Abstract

Community detection in social networks is a cornerstone of social network analysis, yet traditional methods often overlook the intricate interaction patterns between nodes, focusing instead on topological measures such as modularity or centrality. To address these limitations, this paper presents ARLClustering, an R package that employs association rule mining (ARM) through the Apriori algorithm to identify communities based on frequent interaction patterns. By leveraging ARM, ARLClustering uncovers nuanced relational structures, enabling the detection of meaningful and interpretable communities in complex networks. The used methodology involves transforming network data into a transactional format suitable for ARM, identifying frequent itemsets, and extracting communities from association rules. This approach is validated across diverse datasets (e.g. Karate Club, Dolphins, and Facebook networks and so on). A Comparative analysis highlights ARLClustering's important performance in detecting granular clusters and managing high volumes of network data while maintaining computational efficiency. Key contributions include its open-source accessibility, user-friendly design, and robust handling of parameter sensitivity for support and confidence thresholds. Beyond theoretical advancements, ARLClustering provides a practical R package for researchers, with comprehensive documentation and visualization features enabling comparative insights with traditional methods. Future work will enhance its visualization techniques, refine cluster evaluation metrics, and expand real-world applications to broaden its impact. Published on CRAN and available on GitHub, ARLClustering stands as a significant step forward in association rule-based community detection, bridging the gap between traditional methods and the demand for deeper relational insights in social network analysis.