Community Search in Attributed Networks using Dominance Relationships and Random Walks

TL;DR

This work addresses local community search in attributed networks by balancing topology with node attributes through a domination-score framework combined with $k$-core extraction. It introduces two complementary methods, a hop-based approach (HBA) and a random-walk-based approach (RWBA), to identify query-containing cohesive subgraphs and evaluates them on a large real-world dataset. Theoretical analyses derive the computational complexities of both methods, while extensive experiments demonstrate that RWBA can achieve comparable community quality to a 2-hop baseline with substantial runtime benefits under moderate parameter settings. The results highlight a practical, scalable pathway for attribute-aware community discovery in large networks, with implications for social networks and recommender systems.

Abstract

Community search in attributed networks poses a dual challenge: balancing structural connectivity -- the network's topological properties -- and attribute similarity -- the shared characteristics of nodes. This paper introduces a novel algorithm that integrates hop-based and random-walk-based methods to identify high-quality communities, effectively addressing this balance. Our approach employs the concept of the domination score to quantify the influence of nodes based on their attributes, followed by $k$-core extraction to ensure strong structural cohesion within the communities. By considering both the network structure and node attributes, the algorithm identifies communities that are not only well-connected, but also share meaningful attribute similarities. We evaluated the algorithm on large real-world datasets, demonstrating its ability to efficiently identify cohesive communities, making it suitable for applications such as social network analysis and recommendation systems.

Figures (15)

• Figure 1: Example of community search: the query node is highlighted in red. The different subgraphs that may be defined are shown in the dashed circular areas. The selected subgraph is shown in the bottom right corner.
• Figure 2: Example of HBA result
• Figure 3: Example of random walk-search algorithm
• Figure 4: Example of the Grid that is used to calculate the domination score of node $n_x$. Cell $C$ is completely dominated while nodes that are included in cells $A$, $F$, and $E$ need to be evaluated
• Figure 5: Average similarity of the communities produced after running variations of the random walk algorithm compared with the 2-hop search algorithm

Theorems & Definitions (2)

• Definition 1: dominance
• Definition 2: domination score