Comparision of exchange market predictions using extremal data

Abstract

High and low data otherwise then close and open data are not adventitious on time serias curve. Its are extremes and very interesting for traders. Our model based on Evolino RNN ensemble give two distributions based on high and low data. Composition and parametres of this distributions determine the decision of trading. In this paper portfolio constructed by this new method of prediction is compared with portfolio based on extreme moving average method. Different measures of portfolio efficiency and prediction accuracy was used to judge the new prediction method.