A fuzzy adaptive resonance theory—supervised predictive mapping neural network applied to the classification of multivariate chemical data

Abstract

A fuzzy adaptive resonance theory—supervised predictive mapping (Fuzzy ARTMAP) neural network has been studied for the classification of multivariate chemical data. Fuzzy ARTMAP achieves a synthesis of fuzzy logic and adaptive resonance theory (ART) by exploiting the close formal similarity between the computations of fuzzy subset membership and ART category choice, resonance, and learning. To examine the properties of Fuzzy ARTMAP, the well-known Italian olive oil data set was employed. Then this method was applied to a practical agricultural data set to classify different soil samples depending on the crops grown on them. For comparison, the back-propagation (BP) neural network has also been used to treat these data. The results show that the classification performance of the Fuzzy ARTMAP neural network is as good or better than the BP network in the present applications. Among other features, the Fuzzy ARTMAP needs less training time and fewer algorithmic parameters to be optimized than BP does to achieve good classification.