The Classification of Red Wine Quality Based on Machine Learning Techniques

Abstract

The main purpose of this research is to observe the relationships between physicochemical properties (inputs) and sensory (output) variables in the practice of data exploring and analysis. Primarily, this paper investigates the correlation between physicochemical properties with the red wine quality. After fundamental analysis, the classification models of the red wine quality are constructed including KNN, XGB, SVC, and Random Forest models. Based on the evaluation metrics, the Random Forest model has the highest accuracy eventually. It is concluded that the concentrations of sulphates and alcohol have positive influences on red wine quality while lowering the concentration of volatile acidity can also increase the quality. According to the outputs of the models created, the Random Forest produced the best performance according to the accuracy, precision, and f1-score values. The main purpose of this paper is to perform scientific observations of red wine quality based on 11 physicochemical data. These results shed light on the physicochemical that are having a positive correlation with the increment of red wine quality, which provides instrumental suggestions for producing high-quality red wine. © 2023 SPIE.