Evolutionary Discovery of Co-Movement Patterns Among Foreign Currencies

Changes in currency exchange rates can bring significant impact to international macroeconomics and vice versa. Discovering co-movement patterns of currencies in a specific period can help explain economic and financial relationships between the countries under the flexible exchange rate system. In order to discover various co-movement patterns of foreign exchange rates, we propose an evolutionary biclustering algorithm to search for the biclusters each of which has a subset of rows (time points) and a subset of columns (currencies), indicating a group of currencies co-moved within a specific period. We test this algorithm on the historical monthly data of 15 currencies, and successfully detect a number of co-movement patterns. We also find that the discovered patterns will be useful as a guide for foreign currency investment.