Predicting exchange rate is always an interesting issue for both economic and academic communities. The power of forecasting exchange rate accurately, could provide considerable benefits to both firms and investors. But fluctuations in exchange rate, which is caused by various parameters effective in market, have made this job very complex and risky. Until now various methods from economic techniques to pattern recognition techniques in past data have been used in this area of research. One of the methods which has been very popular in last two decades is soft computing. In this research, a hybrid neuro-fuzzy system based on interval type-2 fuzzy c-means clustering, MLP neural network and interval type-2 fuzzy model is proposed for predicting the noisy forex market. To gain faster convergence in learning procedure, combination of back-resilient and back-propagation is used. Two EURUSD and USDCHF exchange rates from forex market are used for experiments. The model is tested for convergence speed and accuracy of prediction. It is also compared with its fuzzy c-means based type-1 equivalent and a FLANN based neuro-fuzzy system. The performance of proposed model in convergence speed and prediction accuracy is proved by experimental results. Keywords: Exchange rate prediction, Neuro-fuzzy system, IT2 fuzzy, IT2 fuzzy c-means
