PERAMALAN NILAI TUKAR MATA UANG MENGGUNAKAN HIBRIDISASI EXPONENTIAL SMOOTHING DAN BACKPROPAGATION NEURAL NETWORK

Abstract

ABSTRACT

This study aims at finding out the use of hybridization of exponential smoothing and backpropagation neural network methods to improve prediction of financial timeseries data, so that the result could contribute useful information to investors in taking important decisions. In financial time series data design, linear and non linear models are often found simultaneously.Thus, using one model is not enough since it may cause another model does not appear. The combination of these two models or hybridization of linear and non linear models are needed to overcome the problem. This method is used to predict the exchange rate of Indonesian Rupiah (IDR) toward US Dollar and Saudi Arabian Riyal. The first thing in conducting this study was predicting the data using exponential smoothing method and continued by backpropagation neural network method. The result of the prediction were incorporated into a combined module in order to reach a result that could generate synergies as the final result. In conclusion, the result showed that hybridization method performs better than early method and this method could act as one alternative in predicting currency exchange.

Keywords: Currency Exchange Rate, Hybridization, Exponential Smoothing, Backpropagation Neural Network.