Community Structures in Trade Flow

Trade flow has been found to be correlated with war; that is, the more trade there is between nations, the less likely they are to go to war with each other. We explore further graph analysis on a data set of trade between countries originally compiled to analyze how trade influences war alliances and country stability. What was missing from this study was seeing if war trends could have been surfaced from the trade graph itself. Here we approach using cluster coefficients, shortest paths, spectral clustering, and graph evolution to study patterns in the trade dataset. Our end conclusion was that approaches that take into account weighted directed edges to separate the graph would give better insights to possible partitions in the data that might resemble alliances or communities of countries. At best, we were able to find a community of war-stable European countries using spectral clustering. Keywords—War, trade, clustering, spectral clustering, graphs