Pemodelan Kurs Rupiah Terhadap Dollar Amerika Serikat Menggunakan Regresi Penalized Spline Berbasis Radial

Abstract

Exchange rate is the price of a currency from a country that is measured or expressed in another country's currency. A country's currency exchange rate has fluctuated due to exchange rate determined by the demand and supply of the currency. One of method that can be used to predict the exchange rate is the classical time series analysis (parametric). However, the data exchange rate that fluctuates often do not fulfill the parametric assumptions. Alternative used in this research is penalized spline regression which is nonparametric regression and not related to the assumption of regression curves. Penalized spline regression is obtained by minimizing the function Penalized Least Square (PLS). To handle the numerical instability and changing data then used radial basis at Penalized spline estimator. Selection of the optimal models is rely heavily on determining the optimal lambda and optimal knot point that is based on the Generalized Cross Validation (GCV) minimum. Using data daily exchange rate of the rupiah against the US dollar in the period of June 2, 2014 until February 27, 2015, the optimal penalized spline bases on radial model in this study is when using 2 order and 13 knots point, those points are 11625; 11669; 11728; 11795; 11911; 11974; 12069; 12118; 12161; 12372; 12452; 12550; 12667 with GCV = 3904.8. Keywords: exchange rate, penalized spline, radial bases, penalized least square, generalized cross validation