Evaluation model of grape wine quality based on BP neural network

In order to select the better wine grape varieties, and improve wine quality evaluation standards, the paper made a cluster analysis of wine grape samples based on the selected 57 physicochemical indexes. It's also classified different categories of wine grape after comparing the results of the wine quality evaluation. SPSS for canonical correlation analysis was used to find typical indicators, which reflect the comprehensive level of wine grape and wine physicochemical indexes. Matlab was used to establish BP three layer neural network, which reflects the relationship among the comprehensive score of wine and typical indexes of grape wine and wine grape. Through the analysis and test on the model, consistent rate reached 100% which proves the model to be available. So we can use the model with physical and chemical indexes of the grape and wine to quantify the quality of wine according to the corresponding training function.