Prediction in complex systems: The case of the international trade network

• The international trade data are represented by a country-product network.
• Diffusion on a network is used to predict the future links.
• Different ways to alter the original diffusion scheme are proposed.
• Diffusion performs best when coupled with a time-aware metric for product similarity.
• The investigated methods can be applied to any network with time information.

Predicting the future evolution of complex systems is one of the main challenges in complexity science. Based on a current snapshot of a network, link prediction algorithms aim to predict its future evolution. We apply here link prediction algorithms to data on the international trade between countries. This data can be represented as a complex network where links connect countries with the products that they export. Link prediction techniques based on heat and mass diffusion processes are employed to obtain predictions for products exported in the future. These baseline predictions are improved using a recent metric of country fitness and product similarity. The overall best results are achieved with a newly developed metric of product similarity which takes advantage of causality in the network evolution.