Identifying Geographic Clusters: A Network Analytic Approach

TL;DR

This paper introduces a novel network-based methodology using k-shell decomposition to identify and analyze geographic clusters of scientific research activity, providing a flexible alternative to traditional fixed-region approaches.

Contribution

It presents a new graph approach for detecting geographic clusters of human activity, applicable across various empirical settings, demonstrated on biomedical research data.

Findings

Identifies consistent research clusters across different geographic scales.

Highlights top research institutions within each cluster.

Reveals spatial patterns of scientific productivity.

Abstract

In recent years there has been a growing interest in the role of networks and clusters in the global economy. Despite being a popular research topic in economics, sociology and urban studies, geographical clustering of human activity has often studied been by means of predetermined geographical units such as administrative divisions and metropolitan areas. This approach is intrinsically time invariant and it does not allow one to differentiate between different activities. Our goal in this paper is to present a new methodology for identifying clusters, that can be applied to different empirical settings. We use a graph approach based on k-shell decomposition to analyze world biomedical research clusters based on PubMed scientific publications. We identify research institutions and locate their activities in geographical clusters. Leading areas of scientific production and their top performing research institutions are consistently identified at different geographic scales.