Foreign Exchange Rate Forecasting Using Improved Artificial Neural Networks in Incomplete Data

In financial markets and specifically exchange rate markets, the environment is fun of uncertainties and changes occur rapidly. Therefore, forecasting in thesesituations requires methods that also work efficiently with incomplete data. In this paper, an improved version of artificial neural networks is proposed by applying the fuzzylogic in order to yield more accurate results, especiaUy for cases where inadequate historical data are available. In our proposed model, instead of using crisp connected weights in traditional artificial neural networks, they are considered as fuzzy numbers to reduce the required data in the training process and improve the performance of the proposed model, especially with scant data. The empirical results of exchangerate forecasting indicate that the proposed model can be an effective way to improve the forecasting accuracy, especially in incomplete data situations.

Keywords:

Artificial Neural Networks (ANNs) , Fuzzy logic , Time series forecasting , Hybrid models , Exchange rate