The Prediction Of The Exchange Rate With Genetic Algorithm Combined With Least Square Method

With the further development of International Economic Integration, Exchange rate in economic activities play an prominent role。There are lots of factors affecting the fluctuation of exchange rates, such as: politics economic, besides, Even including the unexpected natural disaster (earthquake、tsunami......)。Forecasting the tendency of the exchange rate is an important and tough mission because of all kinds of complicated factors. Comparing with the traditional methods, Genetic algorithm combined with least square method has its unique advantages and features in dealing with this issue。This paper aims to apply Genetic algorithm to the forecasts of the exchange rate and propose the new curve fitting method based on discussing the relationship between exchange rates and interest rates, at last making use of MATLAB and C#achieves this process.Taking consideration of the features of the exchange data, which is numerous、nonlinear and so on. Therefore, sorting out all the data is the most important thing of all, we should transfer the data in excel to database in system by performing batch processing and then handle noisy data in database. Through the above process, we can select the data in any area and any time frame to reduce the workload. This paper work out the fitting curve about relationship between exchange rates and interest rates by using genetic algorithm combined with least square method. Firstly, we use MATLAB to draw scatter diagrams according to the exchange rate data and other economic data, then observing the scatter diagrams to model their relationship。Secondly, we take out data from the database of system and handle them。Finally, predicting the fluctuation of exchange rates with the help of genetic algorithm combined with least square method.For the genetic algorithm, the selection of the encoding is not only depended on the arrangement about individual genes, also on the difficulties and perplexities of search in genetic algorithm combined with least square method. In this paper, we use binary coding in encoding the factorial range of Polynomial, and choose1/r(?)2as adaptive value. Then we can carry out these actions such as copying、exchanging according to the adaptive value. At last, we can draw the relation curve between exchange rates and interest rates to predict exchange rates.