Multi-criteria decision making offers new route to community detection in complex networks

Researchers have proposed a new framework for detecting communities in multi-relational networks, where the same set of entities is connected through several different types of relationships. The work, published in the International Journal of Data Science and Analytics, treats the multiple relations in such networks as multiple criteria and applies a multi-criteria decision making approach to decide how users should be assigned to communities.

Multi-relational networks are common in the real world. The authors cite heterogeneous social networks, protein–protein interaction networks and bibliographic networks as examples. Detecting communities in such networks is described as more challenging than in single-relational networks, because information from the different relationship types must be integrated rather than analysed in isolation.

The proposed framework is organised into four phases: data modeling, seed selection, community assignment and leader update. In the data modeling phase, the multiple relations are represented as multiple criteria. The seed selection and leader update phases are designed to accommodate different centrality measures, while the community assignment phase can employ different similarity measures. This flexibility means the framework provides scope for practitioners to choose the measures best suited to a given dataset, rather than fixing a single algorithmic choice in advance.

By using information about all relations, the framework allocates a user to a particular community rather than relying on any single relationship type alone.

To evaluate the approach, the authors conducted what they describe as an extensive experimental study on a complex real-world multi-view Twitter dataset. The dataset allowed them to test the framework on a network where entities are connected through multiple views or relationship layers, mirroring the heterogeneous structure the method is intended to handle.

The results were compared against baseline methods for the community detection task. According to the abstract, the proposed approach achieved improved scores with respect to three evaluation metrics: NMI, Coverage, and Performance.

The study addresses a recognised gap in network analysis. As the authors note, community detection in multi-relational networks is a challenging task compared with single-relational networks, precisely because it requires integrating information from different types of relationships.

The framework’s modular structure is a notable feature of the work. Because the seed selection and leader update phases accept different centrality measures, and the community assignment phase accepts different similarity measures, the method functions as a general scaffold into which alternative algorithms can be substituted. The authors position this as giving scope to employ different measures across the pipeline, which may make the framework adaptable to datasets with varying relationship structures.

The paper, a regular article in the journal’s volume 22, appears under the title describing a novel framework for community detection in complex networks using a multi-criteria decision making approach. Full access to the complete article, including the detailed experimental results, is provided through the publisher’s subscription channels.

Subject of Research: Community detection in multi-relational complex networks using a multi-criteria decision making framework

Article Title: Novel framework for community detection in complex network using multi-criteria decision making approach

Article References: Singh, D., & Verma, A. (2026). Novel framework for community detection in complex network using multi-criteria decision making approach. International Journal of Data Science and Analytics, 22(1), Article 354. https://doi.org/10.1007/s41060-026-01340-1

Image Credits: AI Generated

DOI: 10.1007/s41060-026-01340-1

Keywords: community detection, multi-relational networks, multi-criteria decision making, complex networks, social network analysis, Twitter dataset, centrality measures, similarity measures, NMI, heterogeneous networks, Novel, framework