Community Detection and Analysis of Political Alliances in the Brazilian Congress Voting Network

Context: Network analysis in legislative studies reveals hidden alliances and ideological shifts, helping to understand evolving party behavior and dynamics in democratic systems. Problem: Many voting network studies lack filtering for non-polarized propositions, weakening ideological clarity and modularity. Without edge pruning, irrelevant connections clutter networks, reducing accuracy in detecting true political alliances and dynamics. Solution: This research uses the Leiden algorithm on polarized propositions with edge pruning to detect cohesive communities in Brazil’s Congress, enhancing modularity and revealing true political dynamics. IS Theory: This study uses Social Network Theory to examine voting alliances and Institutional Theory to analyze how norms shape party behavior in Brazil’s Congress, offering insights into political behavior within complex systems. Method: This study uses a quantitative, descriptive approach with network analysis to examine Brazilian congressional voting patterns. Using public data and the Leiden algorithm, it identifies political communities, enhanced by filtering polarized votes and preprocessing to clarify network structure. Summary of Results: Our analysis of Brazilian congressional alliances showed a 10.95% modularity improvement using the Leiden algorithm with edge pruning, revealing clear ideological divides and a dynamic "swing" community influencing alliances in key years. Contributions and Impact on IS: This research advances IS by using network analysis on political voting to clarify alliances. Techniques like polarized data filtering and network optimization provide valuable tools for analyzing complex networks in IS.