The Accuracy Rate of Holt-Winters Model with Particle Swarm Optimization in Forecasting Exchange Rates

Exchange rates forecasting is a crucial and challenging task. Accurate forecasting of the imminent movements of exchange rates is very important in investments, trade and economics. In this paper, an exponential smoothing using the Holt-Winters Model is used for forecasting exchange rates. Parameter search for the smoothing constants is done through computer simulation using Particle Swarm Optimization (PSO). Experiment results show that PSO is able to compute good values for the smoothing constants, producing forecasts with accuracy in determining the direction (rise or fall) of exchange rates. Furthermore, the computed Mean Absolute Deviation (MAD) and the Residual Standard Error (RSE) of the exchange rate forecasts from the actual observed data indicate that PSO can also be used to improve forecasting precision.