A Prediction Model for Quality of Red Wine through Explainable Artificial Intelligence

In view of the complexity and low efficiency of the wine quality prediction process, this paper compares the accuracy of 7 different Artificial Intelligence (Al) classification models to predict the quality of wine, and find out the accurate and objective quality prediction method. ROC curve, sensitivity, and specificity are used to analyze the good classification performance of this model. Experimental findings indicate that the accuracy rate of red wine quality prediction using a random forest algorithm can reach 76.22%. In addition, three ex-post interpretation techniques are used to analyze the corresponding quality effects of 11 characteristics of red wine, so as to find out the important characteristics that affect the quality of red wine. The ex-post interpretation techniques could provide a certain reference for wine quality inspectors in the evaluation of decision-making, and has a certain significance to promote the production of high-quality wines.