Currency Exchange Rate Prediction for United States Dollar to Malaysian Ringgit with Unscented Kalman Filter Technique

Abstract

The paper focuses on predicting the exchange rate between the United States Dollar and the Malaysian Ringgit (USD/MYR) using the unscented Kalman filter (UKF) technique. The UKF technique employs the unscented transform to generate a set of sigma points and to associate the prior knowledge of the probability distribution for a random variable by propagating the mean and the variance from a nonlinear function. The UKF technique requires time prediction and output estimation. A logistic map model is chosen as the state equation, and the exchange rate data are set for the output measurement. A mean square error is used to measure the UKF technique's performance in predicting the currency exchange rates. For illustration, logistic map models with different growth rate values are considered. The appropriate logistic map model gives the smallest mean square error value after the simulation. Moreover, trial models with a set of growth rate values are examined to find a well-performed logistic map model that can be applied to the UKF technique. The testing results show that the logistic map model provides an accurate prediction solution. In conclusion, the UKF technique is an efficient prediction approach for the prediction of currency exchange rates.