Feature Ranking and Topology of the Foreign Exchange Market

This study employs the feature ranking network method to investigate the foreign exchange (FX) market to uncover the underlying structural transition by observing the dependencies and stability of currencies. For this purpose, the FX market’s time series of 50 currencies is examined from January 2020 to October 2023 against the US dollar, covering the COVID-19 pandemic and the Russia–Ukraine war. Using the random forest regressor, the feature ranking matrix is determined by utilizing the returns of currencies on a given day to predict the feature ranks for the following day. The dependency network is constructed using the threshold method, revealing that the topological properties of the networks undergo significant changes, especially during the war. Asian currencies grab the central positions of the dependency network, indicating their high reliance. We select four representative currencies to provide a clearer and more focused analysis of currency dependency, stability, and entropic trends. It is observed that the war triggers instability in currencies and increases the developing countries’ currency dependence. The global entropy increases with minor fluctuations during the war, and a sharp decline in entropy was observed at the beginning of 2023, indicating an extremely high dependence of the currencies of Russia (RUB), the Philippines (PHP), and Bangladesh (BDT) on others. For comparative analysis, we discuss the topological properties of the EUR-based network alongside those of the USD-referred market. The proposed dependency network–based analytical framework provides valuable and sustainable insights for observing currency resilience and contagion in pandemic and geopolitical events.