Community Detection Algorithms: quantitative and qualitative analysis

This work explores the application of six community detection algorithms - Leiden, Louvain, Girvan-Newman, Label Propagation, Walktrap and Infomap - in three databases extracted from the Overton.io platform. The main objective is to compare the performance of these methods through quantitative analysis, through the modularity metric, and qualitative analysis, considering the thematic coherence of the generated clusters. The results demonstrate variations between the algorithms, both in the values of modularity and in the semantic organization of the communities. Among the methods evaluated, the Leiden algorithm stood out for presenting the highest modularity values for the three bases, in addition to producing more coherent partitions from a thematic point of view.