An evolutionary ensemble-based approach for exchange rate forecasting

Time Series Forecasting is an area being concerned by the researchers. Recently, many time series forecasting solutions have been given, but the forecasting accuracy of these solutions needs to be taken into consideration. In this paper, we propose an evolutionary ensemble-based model and experimented it on four sets of test data on exchange rates (USD, NZD, USD and YEN) against the AUD. We applied differential evolution for finding the set of neural networks that have the best fitness value, we also used backpropagation for the local search. The results showed that ensemble learning model was able to generate the better forecasting results than using ANN with BP or ANN with DE only.