一种基于最小二乘支持向量机的葡萄酒品质评判模型

An Evaluation Model of Wine Quality Based on Least Square Support Vector Machine

对源自UCI数据库的葡萄酒数据进行预处理,选取径向基函数作为最小二乘支持向量机的核函数;然后,根据"一对一"算法设计出最小二乘支持向量机多元分类器,并应用交叉验证算法对参数寻优,建立葡萄酒质量评判模型.同时,用BP神经网络、标准支持向量机分类器对葡萄酒进行训练.对比实验结果表明:最小二乘支持向量机比BP神经网络、标准支持向量机的平均分类准确率高,最高分类准确率为100%.

In this paper,the wine dataset from UCI databases is preprocessed and radial basis function is adopted as the kernel function of least square support vector machine（LS-SVM）.And then a multi-classifier is designed from LS-SVM according to one-against-one algorithm.In addition,the cross-validation method is used to optimize parameters and the wine quality evaluation model is built.Meanwhile,LS-SVM is used in the wine quality evaluation and compared with the evaluation methodology based BP（back propagation） neural network and standard support vector machine.Simulation results show that the LS-SVM can achieve higher accuracy than BP neural network and standard support vector machine,with a highest 100% rate.