Informasi Detil Paper

Judul: Applied of Backpropagation Algorithm to Analyzing and Forecasting of Currency Exchange Rate Rupiahs and Dollar
Penulis: Dorteus Lodewyik Rahakbauw  || email:
Jurnal: Prosiding FMIPA 2015 Vol. 1 no. 1 - hal. 99-108 Tahun 2015  [ MIPA ]
Keywords:  Backpropagation, Artificial Neural Networks, Exchange rate
Abstract: Exchange rate or currency exchange rate is very important in the economy. There are three kinds of exchange rate type, namely the selling rate, buying rate and the middle rate. The exchange rate is needed to determine something that needs to be done with regard to the exchange rate eg short-term investment decisions, capital budgeting decisions, long-term financing decisions, and profit assessment. Therefore, it is necessary to attempt forecast the magnitude of the exchange rate for some time to come. The problem faced is how to predict the magnitude of the exchange rate that produces the predicted value with minimal error rate. Forecasting is a process to predict events or changes in the future. In a process activity, the forecasting process is the beginning of a series of activities, and as a starting point the next activity. Modeling of time series is often associated with the process of forecasting the certain of characteristic value in the period ahead, to control a process, or to identify patterns of behavior of the system. By detecting patterns and trends in the data, and then formulate a model, then the data can be used to predict the future. Models with high accuracy will cause the predictive value valid enough to be used as a support in the decision making process. One of the forecasting method developed at this time is using Artificial Neural Network (ANN), which have been the interesting object of ANN research and widely used to solve the problem in some areas of life, one of which is for the analysis of time series data, on the problem Forecasting (Loh, 2003). One of the networks that are often used for the prediction of time series data is Backpropagation neuron network. In this research will be discussed on the use of back propagation neural network to predict the selling rate Rupiah (IDR) per 1 US Dollar (USD). In this study will be shared as much as 70% of existing data as training and 30% of the data as the test data. And in this study used the data of exchange rate in October 2013-January 2014, which is taken from Bank Indonesia site. In research process, Learning rate that used for daily data is 0.5, the process stops at iteration 27088 for daily data, with the gradient achievement is 0.0081822 and the value of R for the training data is 0.99494 which means very good. Furthermore, the data in the test and obtain R value is 0.48638, which means still said to be good in forcasting test data. Some things that affect the results of the research is historical data used for variable ANN enter a lot less, the data used to predict the exchange rate can not be represented as the main factors affecting the exchange rate, and less value of error boundary and suitability weights in the network architecture.
File PDF: Download fulltext PDF PDF

<<< Previous Record Next Record >>>