Analysis of Hybrid Soft and Hard Computing Techniques for Forex Monitoring Systems

Abstract: In a universe with a single currency, there would be noforeign exchange market, no foreign exchange rates, and no foreignexchange. Over the past twenty-five years, the way the market hasperformed those tasks has changed enormously. The need for intelligent monitoring systems has become a necessity to keep trackof the complex forex market. The vast currency market is a foreignconcept to the average individual. However, once it is broken downinto simple terms, the average individual can begin to understand the foreign exchange market and use it as a financial instrument forfuture investing. In this paper, we attempt to compare theperformance of hybrid soft computing and hard computingtechniques to predict the average monthly forex rates one monthahead. The soft computing models considered are a neural networktrained by the scaled conjugate gradient algorithm and a neuro-fuzzy model implementing a Takagi-Sugeno fuzzy inference system.We also considered Multivariate Adaptive Regression Splines (MARS), Classification and Regression Trees (CART) and a hybridCART-MARS technique. We considered the exchange rates ofAustralian dollar with respect to US dollar, Singapore dollar, NewZealand dollar, Japanese yen and United Kingdom pounds. Themodels were trained using 70% of the data and remaining was usedfor testing and validation purposes. It is observed that the proposedhybrid models could predict the forex rates more accurately than allthe techniques when applied individually. Empirical results alsoreveal that the hybrid hard computing approach also improvedsome of our previous work using a neuro-fuzzy approach