Analisis Komparatif Algoritma Infomap, Label Propagation, dan FluidC dalam Deteksi Komunitas Jaringan Undang-Undang Republik Indonesia

Abstract

Setiap undang-undang di Indonesia pada bagian “Mengingat” dalam konsiderans memuat rujukan terhadap undang undang sebelumnya. Seiring dengan terbitnya undang-undang baru setiap tahun, jaringan keterkaitan antar undang-undang menjadi semakin kompleks dan sulit ditelusuri. Untuk itu, diperlukan pendekatan berbasis social network analysis, khususnya deteksi komunitas, guna memetakan dan mengidentifikasi pola keterkaitan tersebut. Penelitian ini mengevaluasi kinerja tiga algoritma deteksi komunitas, yaitu: Infomap, Label Propagation, dan Fluid Communities (FluidC), dalam mengidentifikasi komunitas pada jaringan undang-undang Indonesia periode 2019–2024. Dataset yang digunakan berbentuk graf berarah, di mana simpul merepresentasikan undang-undang dan sisi menunjukkan hubungan rujukan antar undang-undang. Evaluasi algoritma dilakukan menggunakan empat metrik: modularity, coverage, conductance, dan inter-cluster density. Hasil analisis menunjukkan bahwa Label Propagation unggul pada coverage (0,890), conductance (0,331), dan density (0,498), sehingga lebih efektif dalam menangkap kohesi tematik pada jaringan hukum. Infomap dan FluidC mencatat modularity tertinggi (0,433), tetapi menghasilkan komunitas dengan kepadatan internal yang lebih rendah. Berdasarkan temuan tersebut, Label Propagation direkomendasikan sebagai pendekatan yang lebih tepat untuk analisis jaringan undang-undang di Indonesia. In Indonesian legislation, the “Considering” section of each law’s preamble frequently contains references to preceding laws. As new laws are enacted each year, the network of interconnections among these legal documents has grown increasingly complex, making it difficult to trace and analyze their relationships. To address this challenge, a social network analysis (SNA) approach, particularly community detection, is required to map and identify the underlying patterns of legal interrelations. This study evaluates the performance of three community detection algorithms—Infomap, Label Propagation, and Fluid Communities (FluidC)—in identifying communities within the Indonesian legislative network for the period 2019–2024. The dataset is modeled as a directed graph, where nodes represent individual laws and edges indicate citations between them. Algorithm performance was assessed using four metrics: modularity, coverage, conductance, and inter-cluster density. The analysis results show that Label Propagation outperformed the others in terms of coverage (0.890), conductance (0.331), and density (0.498), demonstrating its effectiveness in capturing thematic cohesion within the legal network. In contrast, Infomap and FluidC achieved the highest modularity scores (0.433) but produced communities with lower internal density. Based on these findings, Label Propagation is recommended as a more suitable approach for analyzing Indonesia’s legislative networks.

Keywords

• deteksi komunitas
• jaringan hukum
• infomap
• Label Propagation
• Information System
• Fluid Communities