Evaluation Model Of Wine Quality With Small Annotated Samples

With the rapid development of society and the general improvement of people's living standards,wine is increasingly loved by consumers.In China,the demand for wine is gradually increasing,and major wine manufacturers are paying more and more attention to the Chinese market,and wine quality assessment has become a key element.Wine quality assessment occupies a very important part in the research and development process,which can be used to improve the winemaking process and classify the wine quality.The evaluation of wine quality often uses physical and chemical experimental tests to obtain relevant physical and chemical indicators,such as the determination of acidity,alcohol or p H,while the sensory test mainly depends on the sensory evaluation of experts,such as professional sommeliers.However,the number of high-level sommeliers in China is small,and in the face of the wide variety of wines produced by many producers in the wine industry,relying on only a few professional sommeliers to evaluate wines consumes a lot of manpower and financial resources.Can not achieve the desired effect.Therefore,the difficulty of wine grading lies in data labeling.With the development of machine learning,more and more classification algorithms can be used by us,and support vector machine(SVM)is one of them.This paper takes 4898 cases of white wine data set in the UCI machine learning database as the research object,uses the new active learning multi-classification SVM algorithm and the traditional active learning multi-classification SVM algorithm for horizontal comparison,and proves that the new active learning multi-classification SVM algorithm can be perfectly selected a small number of samples representing the characteristics of the data set are used as the training set.In the case of a small number of training sets,the classification accuracy is significantly better than the other two algorithms.Comparing the new active learning multi-classification SVM algorithm with the traditional Logistic regressionand Ada Boost algorithm,under the same classification accuracy,the new active learning multi-classification SVM algorithm requires a much smaller training set sample size than the other two algorithms.In practical applications,this can greatly reduce the workload of data labeling,make up for the short board that sommeliers are unable to label a large amount of data,and more effectively carry out wine rating.