Foreign Exchange Rate Prediction between Indonesian Rupiah and U.S. Dollar Using Transductive Learning

Purchasing goods or services produced in United States would force Indonesian company or investor to purchase U.S. dollar, and vice versa. The drastically changes of the foreign exchange rate between Indonesian rupiah and U.S. dollar would significantly affect the good’s price. Those facts motivated many studies focused on the exchange rate prediction. Various algorithms have been developed in which data mining has received arising attentions. The aim of this study is to evaluate the transductive learning to forecast the U.S. dollar price from Indonesian rupiah price. Compared to inductive learning, transductive learning is expected to perform better in prediction task. Tomorrow price of U.S. dollar could be difficult to be predicted because the tomorrow price would drastically increase or decrease. Using transductive learning, only small training data set is analyzed to determine the appropriate rule for prediction result. Since the rule is obtained from only small subset of training data, not from the whole, the rule is not generalized as which is applied by inductive learning approach. The result of the small subset of the training data would be used to find the best result of prediction. Index Terms — exchange rate prediction, inductive learning, nearest neighbor, transductive learning.