Prediction of multi currency exchange rates using correlation analysis and backpropagation

In the rapid Development of information and the collecting data collection issue of information network is becoming one of the essential elements affecting many areas, such as foreign exchange (Forex). Forex consists of data having particular ordered values in terms of time history. These values have meaning and can be further predicted for the next value. It is a very important issue of making decision for foreign exchange player (trader) in foreign exchange market. Accurate prediction of forex will give benefit to forex player. But in reality, it is very hard to realize it due to the big piles of data that are necessarily to be processed. This study will develop a system implementing a method so called as Backpropagation (BP) with additional algorithm so called Levenberg Marquardt (LMA) that can predict foreign exchange value, especially for EUR or USD currency. Moreover, input parameter increase will also be developed on BP LMA using Pearson correlation coefficient analysis that will check the correlation between the two variables, but It did not still decrease the error value. The result obtained from conducting testing, forex prediction will be implemented using BP architecture and LMA with best MAPE tryout score of 0. 2208% and best MAPE testing by 0.2693% on scenario 1 (without correlation). In addition MAPE tryout score of 0.3905 % MAPE practicing and MAPE testing score of 0.3816 % on scenario 2 (with correlation). Based on this study we can conclude that using BP LMA still generating better error value than using BP LMA added with correlation analysis on foreign exchange data pair. The margin of BP-LMA and BP-LMA added with correlation analysis is 0.1697 % for MAPE tryout and 0.1123 %. For MAPE testing.