NETW-SCI-EXPLORER

Comparing Clustering Algorithms

Clustering algorithms group similar data points based on their features.

The Louvain Algorithm is an iterative & greedy community detection algorithm that maximizes modularity through local optimization.

Spectral Clustering is an eigenvector-based clustering detection algorithm that uses the graph Laplacian.

The Girvan-Newman algorithm was developed in 2002 and detects clusters based on edge betweenness with hierarchical edge removal ( GN ).

The three are visualized here on the Strike and Karate datasets (see code).

Spectral Graph based on the Laplacian

The OPSAHL-Powergrid U.S Dataset provides insights about U.S power supply lines.

Computing its normalized Laplacian matrix, we use its second and third eigenvalues for node placement, thus visualizing the data in a spectral layout.

The denser central region indicate a core of structurally similar nodes, while the sprawling branches underscore smaller subnetworks, aligning with an expected power-grid network structure.