Abstract
Physical Review E, Vol. 72, No. 4: 046106 (2005) In a system containing a large number of interacting stochastic processes,
there will typically be many non-zero correlation coefficients. This makes it
difficult to either visualize the system's inter-dependencies, or identify its
dominant elements. Such a situation arises in Foreign Exchange (FX) which is
the world's biggest market. Here we develop a network analysis of these
correlations using Minimum Spanning Trees (MSTs). We show that not only do the
MSTs provide a meaningful representation of the global FX dynamics, but they
also enable one to determine momentarily dominant and dependent currencies. We
find that information about a country's geographical ties emerges from the raw
exchange-rate data. Most importantly from a trading perspective, we discuss how
to infer which currencies are `in play' during a particular period of time.
