"VALIDATION AND VERIFICATION OF MONEY MARKET FORECASTING WITH A NON-LINEAR METHOD (USD VS IDR CURRENCY CASE)"
In this paper, non-linear method is used for 5 days forecast of interbank exchange rate for US Dollar (USD) against Indonesia Rupiah (IDR) currency based on October 23, 1993 through February 20, 2004 observed interbank exchange rate daily data. Interbank exchange rate daily anomaly prediction was done for daily interbank exchange rate where here I evaluate anomaly is daily data subtracted by average from observation period. The optimal parameters of non-linear method have been gotten in prediction exchange rate for US Dollar (USD) against Indonesia Rupiah (IDR) currency anomaly were: 2 days lag time, 24 embedding dimensions. The result of model validation was got 0.99 (coefficient correlation). The results model prediction will show that interbank exchange rate anomaly for first next 3 days is very good performance in accurateness and trend but next 2 days is poor in accurateness but good in trend. But if I make average for 7 best embedding dimension the result is opposite (first next 3 days is good in trend only but not in accurateness and the rest of next 2 days is very good performance in accurateness and trend. Therefore, non-linear method is very effective for shorterm prediction in money market, but if we want to make longterm prediction, we must run the model for several dimension embedding and choose some of the best result for the averaged.