Structure Profile by a Model of Precipitation

TL;DR

This paper introduces a precipitation-based model to analyze network structures by grouping nodes into communities without prior knowledge, revealing dense interconnection regions and applicable to directed and weighted networks.

Contribution

It presents a novel precipitation-inspired method for community detection that does not require prior community information and can handle various network types.

Findings

Effectively identifies community regions in networks.

Applicable to directed and weighted networks.

Reveals dense interconnection regions in adjacency matrices.

Abstract

The organizational structure of a network is investigated with a simulated precipitation model which does not make use of prior knowledge about the community structure of the network. The result is presented as a structure profile through which various definitions of communities can be applied for specific applications. The simulated precipitation model performs the grouping of nodes so that nodes belonging to the same 'community' automatically aggregate, thereby revealing regions of the adjacency matrix with denser interconnections. The process is analogous to massive particles precipitating towards the lower potential layer. Without loss of the infrastructure information, a community structure profile of a network can be obtained as the ground state of the Hamiltonian. The method is also applicable to directed and weighted networks.