Forecasting the Exchange Rate Using Meta-Heuristic Algorithms

Adopting appropriate exchange rate policies in developing countries is always controversial. In order to prevent losses from changes in exchange rates, monetary policymakers have always sought to find a suitable method for forecasting exchange rates. However, the political and economic characteristics of the exchange rate have caused complex and nonlinear behavior, indicating the use of better models in forecasting. In this study, the optimal exchange rate model was simulated using metaheuristic algorithms. For this purpose, from 257 monthly data of exchange rate, inflation rate, OPEC basket oil price and gold coin price during the period from December 2000 to March 2021, the data were first divided into two groups of training and testing. Each of the metaheuristic algorithms was run for the next 24 months with the parameters related to each algorithm. In each run, the value of the error coefficients was selected after reaching the stop criterion and finally the best algorithm was selected based on the highest convergence. The results show that the particle swarm algorithm had very good convergence in low number and iteration and had a more accurate performance in forecasting the exchange rate. Given that metaheuristic algorithms are capable of optimizing various processes such as time management, resources, and facilitating daily planning, this study suggests that policymakers use this algorithm to predict the exchange rate before planning in sectors.