The aim of this paper is to help select suitable grapes for making red wine by studying the correlation between ingredients of the grapes and the resulting wine. Two progressive models are carried out in determining the relation between physical and chemical indexes of the grapes and the resulting wine. Multiple linear regression model is used to get the function relationship between them and the related weight. According to the weight, the less relevant indexes of grape are eliminated. Single layer forward neural network is established between the rest physical and chemical indexes of the grapes and wine to make a more accurate solution of the weight. The correlated factors of each index in grape wine are obtained by analyzing the related weight. The anthocyanin, DPPH radical and soluble solid in grapes has a considerable effect on the quality of red wine.