A neural network application in the design of emulsion-based products

In the design of emulsion-based products, consumer appreciation is the key aspect. While the most commonly applied product/process design strategies lack the incorporation of consumer desires, this contribution describes a design approach where consumer appreciation is the main objective. The first step of the design is to translate the consumer needs into quantifiable product attributes, e.g. creaminess and firmness for a mayonnaise. A tasting panel is employed to rate these attributes. A Neural Network (NN) is then applied to correlate these attributes with a characteristic product viscosity. The uncertainty of both the measurements and the panel ratings is included in the training and validation of the neural network. According to the Akaike information criterion NN is inferior to partial least squares regression but NN scores better on the validation test; RMSE = 10%.