[Paper Review] Who is the best connected EC researcher? Centrality analysis of the complex network of authors in evolutionary computation

J. J. Merelo, Carlos Cotta|ArXiv.org|Aug 15, 2007

Complex Network Analysis Techniques13 references3 citations

TL;DR

This paper applies multi-centrality analysis and Pareto dominance to identify the most influential researchers in evolutionary computation (EC) via co-authorship networks. By combining degree, closeness, betweenness, and eigenvector centrality, it identifies K. Deb and D.E. Goldberg as consistently top-ranked, revealing structural hubs and social dynamics in the EC research community through a robust, multi-metric framework.

ABSTRACT

Co-authorship graphs (that is, the graph of authors linked by co-authorship of papers) are complex networks, which expresses the dynamics of a complex system. Only recently its study has started to draw interest from the EC community, the first paper dealing with it having been published two years ago. In this paper we will study the co-authorship network of EC at a microscopic level. Our objective is ascertaining which are the most relevant nodes (i.e. authors) in it. For this purpose, we examine several metrics defined in the complex-network literature, and analyze them both in isolation and combined within a Pareto-dominance approach. The result of our analysis indicates that there are some well-known researchers that appear systematically in top rankings. This also provides some hints on the social behavior of our community.

Motivation & Objective

• To objectively identify the most connected and influential researchers in the evolutionary computation (EC) community using network centrality metrics.
• To overcome limitations of single-metric centrality by combining multiple measures (degree, closeness, betweenness, eigenvector) to reflect diverse aspects of influence.
• To apply Pareto dominance to rank authors holistically, avoiding bias from any single centrality measure.
• To reveal patterns in collaboration dynamics and social structure within the EC research community.
• To provide a benchmark for sociometric influence in EC that complements subjective perceptions of scientific impact.

Proposed method

• Construct a co-authorship network from EC publications, treating each author as a node and co-authorship as an undirected edge.
• Compute four standard centrality measures: degree (number of co-authors), closeness (inverse of average shortest path to others), betweenness (frequency of shortest paths through the node), and eigenvector centrality (influence based on connections to other influential authors).
• Apply Pareto dominance to identify non-dominated authors across all four centrality measures, forming a non-dominated front (Pareto front) of top-ranked researchers.
• Compare results across different combinations of centrality measures to assess robustness and sensitivity of rankings.
• Use visualization and correlation analysis to examine relationships between centrality measures and identify potential biases or redundancies.
• Perform temporal analysis of network evolution to contextualize current centrality rankings within historical collaboration trends.

Experimental results

Research questions

• RQ1 Which EC researchers are most central across multiple network centrality measures, and how consistent are their rankings?
• RQ2 How do different centrality metrics (degree, closeness, betweenness, eigenvector) correlate, and what do their divergences reveal about influence in the EC community?
• RQ3 To what extent does using a multi-objective Pareto dominance approach improve the identification of key researchers compared to single-metric rankings?
• RQ4 How do the top-ranked researchers identified via centrality analysis align with established scientific reputation in the EC field?
• RQ5 What insights into the social and collaborative dynamics of the EC community can be derived from the structure of the co-authorship network?

Key findings

• K. Deb and D.E. Goldberg consistently rank among the top researchers across all four centrality measures and appear in the first non-dominated Pareto front, indicating robust influence.
• The top non-dominated front includes six researchers: K. Deb, H. de Garis, D.E. Goldberg, T. Higuchi, D. Keymeulen, and X. Yao, all of whom are highly recognized in the EC community.
• The second non-dominated front includes prominent researchers such as T. Bäck, C. Coello Coello, D.B. Fogel, and Z. Michalewicz, confirming the method's alignment with expert perception.
• Eigenvector centrality identifies key researchers who may be 'hitchhiking' on the influence of others, but their prominence is reinforced when combined with other metrics.
• The correlation between centrality measures is low when all four are considered together, justifying the use of multi-objective ranking to avoid oversimplification.
• The results suggest that centrality is a dynamic property: top researchers are not only currently influential but have likely been central throughout the network’s evolution.