Exchange rate forecasting by GMDH by Wong Hon Kit Thesis Supervisor: Dr. Simon Fong Foreign exchange market (Forex, FX or currency market) is a decentralized financial market in the world for trading currencies. Except weekend and major holidays, financial centers around the world take turns running by location, so that the foreign exchange market to 24 hours of uninterrupted trading of various currencies. The foreign exchange market determines the exchange rate between different currencies. The exchange rate is the most important international trade adjustment lever. Exchange rate will affect the level of the commodity on the international market, costs and prices, directly affecting the international competitiveness of goods. In recent years, the field of artificial intelligence methods for financial management and forecasting attracted much attention. Forecast financial information, for the prevention of risks and importance of customer transactions, many domestic and foreign institutions of research and development of the exchange rate forecast, relevant research has made great progress. However, fluctuation in exchange rate is affected by a number of uncertainties, it is difficult to find a model that accurately and efficiently characterize the impact of exchange rate fluctuations between the factors and their relationship. Because of the exchange rate market has a data-rich, uncertain, weak model features, so statistical techniques, fuzzy and neural networks, and other nonlinear dynamic modeling of much attention in the research process, and achieved certain results. Analysis of the exchange rate, there are two main methods: fundamental analysis and technical analysis. Based on the analysis of the basic factors that affect exchange rates were analyzed basic factors include the level of economic development of countries and the situation in the world, regional and national political situation, market expectations and so on. Technical analysis is the use of psychology, statistics and other disciplines, research methods and means, through the exchange rate of the previous studies to predict the future trend of the exchange rate. Exchange rate forecasting is a very important subject in financial market. This thesis applies group method of data handling (GMDH) method for exchange rate forecasting. The GMDH algorithm is used to combine the analog patterns and identify an optimum ensemble which has similar characteristics with the modeling object. The empirical results show that this method can well forecast exchange rate.