Multi-attribute community detection in International Trade Network

TL;DR

This paper introduces a novel multi-attribute community detection method for the International Trade Network, combining topological indicators and a fast local search algorithm to identify economically significant clusters of countries.

Contribution

It presents a new multi-criteria approach to weighted network construction and a fast algorithm for clique partitioning, enhancing community detection in trade networks.

Findings

Clusters reflect trade intensity and economic power of countries

Method reveals trade community structures with high intra-group connectivity

Approach improves understanding of international trade patterns

Abstract

Understanding the structure of communities in a network has a great importance in the economic analysis. Communities are indeed characterized by specific properties, that are different from those of both the individual node and the whole network, and they can affect various processes on the network. In the International Trade Network, community detection aims to search sets of countries (or of trade sectors) which have a high intra-cluster connectivity and a low inter-cluster connectivity. In general, exchanges among countries occur according to preferential economic relationships ranging over different sectors. In this paper, we combine community detection with specific topological indicators, such as centrality measures. As a result, a new weighted network is constructed by the original one, in which weights are determined taking into account all the topological indicators in a multi-criteria approach. To solve the resulting Clique Partitioning Problem and find homogeneous group of nations, we use a new fast algorithm, based on quick descents to a local optimal solution. The analysis allows to cluster countries by interconnections, economic power and intensity of trade, giving an important overview on the international trade patterns.