Adaptive neuro-fuzzy inference system (ANFIS) simulation for predicting overall acceptability of ice cream

Abstract

Because of uncertain nature of sensory evaluation due to differences in the individual panelist's perception of the product attributes, application of fuzzy set concept could be useful. In this research, adaptive neuro-fuzzy inference system (ANFIS) was used to predict overall acceptability of ice cream. Consumer acceptance has been recognized as the key driver for product process. Experimental sensory attributes (flavor, body & texture, viscosity and smoothness) were used as inputs and independent overall acceptability as output of ANFIS. Thirty percent, thirty percent and forty percent of the sensory attributes data were used for training, checking and testing of the ANFIS model, respectively. It was found that ANFIS model achieved an average prediction error of overall acceptability of ice cream of only 5.11%. These results indicate that this model could potentially be used to estimate overall sensory acceptance of ice cream and related products.