Predicting the Exchange Rate: Logistic Growth Model and Competing Models

The main objective of this study is to compare the Harvey Logistic Growth Models, Harvey, Nonlinear Autoregressive Neural Network, and to design an optimal model with better predictive accuracy for the exchange rate with high volatility and a nonlinear motion patterns which has been neglected for predicting the exchange rate in Iran.  In this study, we use the "Harvey Logistic" Growth Models, Harvey and adding a nonlinear component based on the Taylor series expansion for trigonometric functions, and using the daily data during 1392:1-1398:3, the fluctuations of the exchange rate and the accuracy and prediction of these models are compared its results with the nonlinear autoregressive neuronal network. The results of unit root tests represent that the data is stationary and has nonlinear property. In the estimation stage, the goodness of fit for the Logistic and Harvey models are not confirmed. By adding nonlinear parts to the Harvey model, a good fit was obtained for the exchange rate with a coefficient of determination about 99.95 percent and a minimum mean square error, even when compared with the nonlinear autoregressive neural network. The results show that combining the Harvey model with the nonlinear component is considered as one of the best models which predicts the exchange rate better than other models.