Hierarchical Community Detection and Network Analysis: A 2-Level Approach for Social Media Networks

As social networking site users and their relationships generate enormous amounts of data, these networks become complicated, requiring sophisticated research approaches to extract valuable information. This research presents an innovative 2-level hierarchical method for community discovery. Greedy Modularity and Label Propagation community detection algorithms applied in the first level to find communities. Next, node similarity is assessed using similarity metrics such Jaccard, Cosine, Katz, and Adamic-Adar. These nodes are grouped utilizing K-Nearest Neighbors (KNN) using common neighbors and vicinity to centroids once initial communities are identified, which refines the neighborhood framework to form a 2-tiered hierarchy. NetworkX visualizations for link prediction techniques used to highlight additional network topology insights. The proposed method improves on conventional community detection by identifying intricate hierarchical patterns, providing a more thorough comprehension of the dynamics of social media networks. The effectiveness of the presented hierarchical community detection methodology is assessed using quantitative indicators, such as F1 score, precision, and recall. The outcomes show the reliability of our approach in accurate community detection.