Prediction and Accuracy of Rupiah Exchange Rates Against US Dollar Uses Radial Basis Function Neural Network

Currency exchange rates, or often referred to as the exchange rate, are the price of one unit of a foreign currency in a domestic currency or can also be called the price of a domestic currency against a foreign currency. The value of a country's currency is strongly influenced by the flow of capital between countries. The high exchange rate of other countries' currencies against a country will result in the deterioration of the economic situation of a country. The weakening of the currency exchange rate will cause Indonesia's foreign debt to increase and the balance sheets of companies and banks will decline. The phenomenon of volatile rupiah exchange rate fluctuations often occurs in Indonesia which will cause economic conditions, especially trade will be disrupted because trade is valued in US dollars (USD). Therefore, serious handling is needed in the face of erratic exchange rate fluctuations because it will affect the economic performance of a country so that a decision can be made after knowing the exchange rate of the next period. In helping to make decisions, the authors make a forecasting model of the rupiah exchange rate against USD using the radial basis function of the neural network. In this research used factors that influence the fluctuation of the rupiah exchange rate against IDR, namely the value of exports, imports, GDP, BI interest rates, inflation rates, and the money supply. In this research optimization of learning rate and hidden neuron parameters was done to get the lowest error value or error rate. The results of the research using the radial basis of neural network functions produce accuracy values ranging from 89 - 95% in the training process while the testing process ranges from 67 - 98% and with an error rate of 4 - 11% in the training process while 2 - 32% for testing process. Key Words : Exchange rates, forecasting, RBF, training, testing, accuracy