Prediction of Foreign Currency Exchange Rates Using CGPANN

This paper contributes an application of Cartesian Genetic Programming Evolved Artificial Neural Network (CGPANN) for forecasting the foreign currency exchange rates. The end product of our work is an efficient Artificial Neural Network (ANN) based prediction model that forecasts the foreign currency exchange rates, making use of the trends in historical data. These trends in the historical currency data serve as significant prognostic factor to train the prediction model. The algorithm exploited for the evolution of the prediction model is Cartesian Genetic Programming (CGP). CGP evolved ANNs have great potential in prediction models for forecasting systems. Historical daily prices of 500 days data of US dollars are monitored to train the prediction model. Once the model is trained, it is tested on 1000 days data of ten different currencies to predict these currency rates and the results are monitored to analyze the efficiency of the system. The results show that prediction model achieved with CGPANN is computationally cost effective and accurate (98.85%) that is unique as it is dependent on least amount of previous data for future data prediction.