基于PCA神经网络的葡萄酒评价模型研究
2016-05-03刘颖王云苗宇雷宋景帅陈丽芳
刘颖 王云 苗宇雷 宋景帅 陈丽芳


摘要:利用主成分分析与RBF神经网络相结合,建立葡萄酒质量评价预报模型,并进行训练和仿真验证。该模型运用SPSS软件对葡萄酒中影响风味指标进行主成分分析,将多变量、非线性的原始数据进行降维,保留原始信息的主要信息,把原来若干个属性变量综合成几个不相关主成分分量;再以计算结果作为RBF网络的输入数据,葡萄酒的感官评价得分作为网络的输出数据,建立葡萄酒主要理化指标与葡萄酒质量的关系模型。结果表明:该评价模型的建立,缩短了葡萄酒评价的周期,克服了品酒师聚集的困难;与传统RBF网络相比,大大简化了网络结构,提高了网络的训练速度和预报精度,为质量评价问题提供了一种的研究思路。
关键词:主成分分析;RBF神经网络;评价模型;理化指标
中图分类号:TP183;O29 文献标识码:B 文章编号:2095-2163(2016)02-
Research of the wine evaluation model based on PCA & neural network
LIU Ying1, WANG Yun2, MIAO Yulei1,SONG Jingshuai1,CHEN Lifang2
(1Yi-sheng innovation Education Base, North China University of Science and Technology, Tangshan Hebei 063009;
2 Science College, North China University of Science and Technology, Tangshan Hebei 063009)
Abstract: Using the principal component analysis combined with RBF neural network, to set up the wine quality forecast model, meanwhile, training and simulating it. By using the SPSS software, realize principal component analysis on the wine flavor indexes, which reducing the multivariable and nonlinear dimension of primary data, retaining the original information of the main information, associating with the original several attribute variables to be synthesized into a few principal components. The calculation results are regarded as input of RBF network, the wine sensory evaluation score as the output of the network, to set up the main physical and chemical indicators with the wine quality relationship model.The results show that, the evaluation model, has shorten the Wine evaluation period, overcome the difficulty of wine taster gathering. Compared with the traditional RBF network, simplify the network structure, improve the network training speed and prediction accuracy, at the same time, provide a research way for quality evaluation.
Keywords:PCA;RBF neural network;Evaluate model;physical and chemical indicators
0 引言
葡萄酒成分复杂,其质量是气味、口感、外观等的综合体现。主要依靠品酒专家进行品评,由其评分来确定葡萄酒的质量[1]。实际生活中,对葡萄酒的评价往往依赖于品酒师的经验,易受到主观因素的影响,而不能如实地反映葡萄酒的实际品质,因此,研究一种客观评价葡萄酒的模型,实现葡萄酒质量的综合、精准评价是非常必要的。本文即以葡萄酒理化指标和品酒师给出的质量评价数据作为研究背景[2],通过主成分分析,对影响葡萄酒品质的因素进行计算、优化及处理,获得预处理后的结果,再结合RBF神经网络完成建模设计。……
