White Wine Quality Prediction and Analysis with Machine Learning Techniques

Abstract

The wine originated 7,500 years ago, and it is one of the most popular alcoholic beverages in the world today. With modern advances in production processes, the quality of wine has been significantly improved. White wines can be graded by measuring various physical and chemical quantities. In reality, it is very rare for white wines to have dramatic quality differences, so precisely capturing these important characteristics will be difficult. In addition, the available dataset for wine quality is relatively small and imbalanced, so it is not easy to train a machine learning model for wine quality prediction. This paper explores the applicability of applying logistic regression machine learning models for wine quality prediction. The Synthetic Minority Oversampling Technique (SMOTE) algorithm is applied to address the imbalanced data issue. The key features in wine production are analyzed to the physical or chemical quantities that are most likely to influence the quality of white wines. In several tests, the random forest model gives good results, and free sulfur dioxide, chlorides, and fixed acidity are the most likely characteristics to influence the quality of white wines.