摘要:The early detection abnormal fermentations(sluggish and stuck) is one of the main problemsthat appear in wine production, due to the significantimpacts in wine quality and utility. Thissituation is specially important in Chile, whichis one of the top ten worldwide wine productioncountries. In last years, two different methodscoming from Computational Intelligence havebeen applied to solve this problem: Artificial NeuralNetworks and Support Vector Machines. Inthis work we present the main results that havebeen obtained to detect abnormal wine fermentationsapplying these approaches. The SupportVector Machine method with radial basis kernelpresent the best results for the time cutoffs considered(72 [hr] and 96 [hr]) over all the techniquesstudied with respect to prediction rates and numberof the training sets.
关键词:Abnormal; Wine Fermentations;Artificial Neural Networks; Support Vector Machines.