SENSORY EVALUATION OF VIRGIN OLIVE OILS BY ARTIFICIAL NEURAL-NETWORK PROCESSING OF DYNAMIC HEADSPACE GAS-CHROMATOGRAPHIC DATA

A different approach to the traditional sensory method was used for th e sensory quality evaluation of virgin olive oils. Two hundred and fou r oil samples differing in their quality, and extracted from olives of various varieties, ripeness, sanitary state and geographical origin, were submitted to sensory evaluation by a panel test and dynamic head- space analysis for the quantification of volatile fractions. An artifi cial neural network (ANN), using the backpropagation algorithm, was ap plied to the head-space results (input) with the aim of predicting pan el test scores (output). It was found that the ANN was able to general ise well and to assign the sensory evaluations with a good degree of a ccuracy. The high proportion of correct answers (96%) suggested that s ensory evaluation from the panel test could be successfully replaced b y the dynamic head-space analysis-ANN coupled approach.