Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quantifying Global Networks of Exchange through the Louvain Method

Aryan Sharma, Jaden Li, Christina Chu, Anna Sisk

TL;DR

This work models global country relations as a weighted network built from Congressional Research Service reports (1996–2024), with 172 countries connected by shared policy topics. It adopts the Louvain clustering method to maximize modularity $Q$ and identify 10 distinct country clusters, while using eigenvector centrality from the adjacency matrix $A$ to rank country influence, highlighting key players such as the United States, Russia, and China. The methodology leverages NetworkX for graph construction and clustering, and provides a reproducible pipeline with publicly available data. The findings offer a framework for analyzing global connectivity and inform evidence sourcing for analytic products, though they acknowledge variability due to the heuristic nature of the algorithm and suggest future enhancements via larger, time-filtered datasets.

Abstract

Congressional Research Service (CRS) reports provide detailed analyses of major policy issues to members of the US Congress. We extract and analyze data from 2,010 CRS reports written between 1996 and 2024 in order to quantify the relationships between countries. The data is processed and converted into a weighted graph, representing 172 unique countries as nodes and 4,137 interests as bidirectional edges. Through the Louvain method, we use a greedy algorithm to extract non-overlapping communities from our network and identify clusters with shared interests. We then compute the eigenvector centrality of countries, effectively highlighting their network influence. The results of this work could enable improvements in sourcing evidence for analytic products and understanding the connectivity of our world.

Quantifying Global Networks of Exchange through the Louvain Method

TL;DR

This work models global country relations as a weighted network built from Congressional Research Service reports (1996–2024), with 172 countries connected by shared policy topics. It adopts the Louvain clustering method to maximize modularity and identify 10 distinct country clusters, while using eigenvector centrality from the adjacency matrix to rank country influence, highlighting key players such as the United States, Russia, and China. The methodology leverages NetworkX for graph construction and clustering, and provides a reproducible pipeline with publicly available data. The findings offer a framework for analyzing global connectivity and inform evidence sourcing for analytic products, though they acknowledge variability due to the heuristic nature of the algorithm and suggest future enhancements via larger, time-filtered datasets.

Abstract

Congressional Research Service (CRS) reports provide detailed analyses of major policy issues to members of the US Congress. We extract and analyze data from 2,010 CRS reports written between 1996 and 2024 in order to quantify the relationships between countries. The data is processed and converted into a weighted graph, representing 172 unique countries as nodes and 4,137 interests as bidirectional edges. Through the Louvain method, we use a greedy algorithm to extract non-overlapping communities from our network and identify clusters with shared interests. We then compute the eigenvector centrality of countries, effectively highlighting their network influence. The results of this work could enable improvements in sourcing evidence for analytic products and understanding the connectivity of our world.

Paper Structure

This paper contains 8 sections, 3 equations, 2 figures, 2 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Background
  3. Louvain Clustering
  4. Eigenvector Centrality
  5. Methodology
  6. Results
  7. Conclusion
  8. Reproducibility Statement

Figures (2)

  • Figure 1: A visualization of our full graph with countries represented by nodes and shared interests represented by edges. The Louvain method forms 10 distinct clusters, each a different color.
  • Figure 2: An arbitrary cluster of our graph that includes Colombia, Sudan, and several other countries that are connected by their common associations with human rights and foreign assistance.