IMPLEMENTASI NUERAL NETWORK BACKPROPAGATION UNTUK MEMPREDIKSI KURS VALUTA ASING

Abstract

Abstract - Technology neural network system has been implemented in various applications, especially in terms of forecasting (forecasting), including backpropagation that can be applied to predict foreign exchange rates. There are two steps being taken in this backpropagation method of training and testing phases. In this network backpropagation algorithm is given a pair of pattern - a pattern that consists of the input pattern and desired pattern. When a pattern is given  to the network, weights - weights modified to minimize the differences in the pattern of output and the desired pattern. This exercise is performed over and over - re-issued so that all the patterns the network can meet the desired pattern. The next stage is the testing stage. This stage begins by using the best weights obtained from the training phase to process the input data to produce the appropriate output. It is used to test whether the ANN can work well is that it can predict the pattern of data that has been drilled with a small error rate. From the test results using data from the monthly period in the training process the network can recognize input patterns are provided so entirely in accordance with the target. While testing with the use of data daily and weekly period that does not comply with the given target, it is because the network requires more data to identify patterns provided. As more data are trained, the better the network will recognize the pattern - a pattern so that the results more accurate predictions, but will be impacted by slowing the process of training.

 

Keywords:  Foreign currency exchange rate prediction, Neural Network, Bacpropagation