The Research On Outlier Data Mining Of The Rmb Exchange Rate Time Series

Exchange rate is not only the main instrument of the macroeconomic regulation, but also an important economic lever which can promote the accomplishment of social development. The abnormal behavior in exchange rate changes will cause more speculation on the foreign exchange market, which will not only badly intervene in the monetary policy and the management of exchange rate, but also create more risks. Since 2005, especially the U.S. subprime mortgage crisis break out triggering the global economic crisis in 2007, the uncertainty contained in RMB exchange rate is more and more significant. Therefore, it is urgent to grasp the RMB exchange rate behavior accurately and strengthening the exchange rate risk management and control furtherly.Based on the discussion of methods for exchange rate behavior study in detail as well as the present research of time series data mining, from a new perspective of "outlier" and by using the data-driven approach, this study tries to explore the information implied in the abnormal change of the RMB exchange rate behavior, which can help to reveal its internal rule. The empirical study included outlier detection and outlier analysis. In the outlier detection part, with the consideration of orderliness and nonlinearities in the exchange rate data, this thesis puts forward a hybrid detective technique by combining the phase space reconstruction with the sequential exception method, which can realistic reflect the RMB exchange rate system characteristics and structural changes; In the outlier analysis part, from the macro perspective, this thesis try to find the interaction mechanism among different factors in the system. With the method of structural equation model, this thesis have found the structure relationship and effect mechanism between the RMB exchange rate and the price level, interest rates, economic growth, the international balance, as well as the monetary policy intervention of the central bank, thereby the thesis provides a reasonable explanation for the generation of outliers.