Resumo. The honey classification depends on the raw material, this classification is important in the appreciatione identification of the product. The characterization can be given by the analysis of physic-chemicalparameters. In this paper, 25 samples of honey were classified into floral honey or honeydew through theKirkwood method. A neural model was developed with multilayer percentron networks, backpropagationlearning algorithm and an easy operation system for the use of beekeepers. Tested networks were able to classify the honey with 100% precision, using only the pH and reducing sugar parameters, which are determinedin an easy, cheap and quick way.