Classification of Colombian honeys by electronic nose and physical-chemical parameters, using neural networks and genetic algorithms

Honey production in Colombia has increased significantly in recent years, but there are few reports related to the quality of this product and its differentiation of its origins. The aims of this study were to determine the physicochemical parameters of honeys produced by Apis mellifera bees in four Colombian regions and to assess the possibility of origin differentiation using chemometric analysis. Two hundred thirty six honey samples were analyzed by 15 standard physico-chemical parameters and by aroma profile using an electronic nose. A genetic algorithm was performed to determine the most influencing variables on the classification of honey, resulting in 14 variables. From this result, an artificial neural network was developed, allowing classification of honey samples according to their geographical origin with a success rate of 81%.