A hierarchical fuzzy system with high input dimensions for forecasting foreign exchange rates

Fuzzy systems suffer from the curse of dimensionality as the number of rules increases exponentially with the number of input dimensions. Although several methods have been proposed for eliminating the combinatorial rule explosion, none of them is fully satisfactory as there are no known fuzzy systems that can handle a large number of inputs so far. In this paper, we describe a method for building fuzzy systems with high input dimensions based on the hierarchical architecture and the MacVicar-Whelan meta-rules. The proposed method is fully automated since a complete fuzzy system is generated from sample input-output data using an Evolutionary Algorithm. We tested the method by building fuzzy systems for two different applications, namely the forecasting of the Mexican and Argentinan pesos exchange rates. In both cases, our approach was successful as both fuzzy systems performed very well.