基于理化指标的BP神经网络葡萄酒质量评价  

Wine Quality Grade System of BP Neural Network-Oriented

作　　者：吴良超[1] 张帆[2] 陈延礼[1] 

机构地区：[1]宜宾学院数学学院,四川宜宾644007 [2]宜宾学院经济与管理学院,四川宜宾644007

出　　处：《宜宾学院学报》2013年第6期43-46,共4页Journal of Yibin University

基　　金：宜宾学院学生科研项目(2012X018)

摘　　要：针对我国葡萄酒业内缺乏利用理化指标对葡萄酒进行评级的现状,分析能否使用葡萄和葡萄酒的理化指标评价葡萄酒的质量.通过运用双因子方差分析、主成分分析、逐步回归分析等方法分析了葡萄酒的分级以及酿酒葡萄与葡萄酒的理化指标之间的联系等问题,建立了基于Matlab平台的BP神经网络模型,得到了在一定条件下,能用酿酒葡萄和葡萄酒的理化指标来评价葡萄酒的质量的结论.但仅考虑理化指标时会使结果存在一定的误差,故建议使用理化指标和简单的感官分析相结合来评价葡萄酒的质量,以提高评价葡萄酒质量的准确性.Aiming at the current situation where there is absence of physical and chemical indexes used to grade wine quality in the wine industry in China, analysis was done to test whether the physical and chemical indexes could be used to grade the wine quality. Based on Matlab platform, the BP neural network model was built after analyzing the connection of the wine grading, and the wine grape and physical and chemical indexes of the wine were graded by using double factor analysis of variance, PCA （ prin- cipal component analysis） and stepwise regression analysis. On certain condition, the wine quality can be graded by using the physical and chemical indexes of wine grape and wine. To improve accuracy, using the physical and chemical indexes combined with simple sensory analysis to grade wine quality is advised, for there may be some errors if only the physical and chemical inde-xes are considered.

关 键 词：双因子方差分析 主成分分析 逐步回归分析 BP神经网络模型 葡萄酒质量 

分 类 号：TS262.6[轻工技术与工程—发酵工程]