Wine Quality Detection Using Machine Learning: A Comparative Analysis of Classification Algorithms

Abstract

The present work describes a fully devel- oped machine-learning based approach for the predic- tion of wine quality from its physicochemical parame- ters. Utilizing the UCI Wine Quality Dataset contain- ing 6,497 observations of Portuguese “Vinho Verde” wine, we apply and contrast various classification techniques such as Random Forest, Support Vector Machine (SVM), XGBoost, K-Nearest Neighbours, and Gradient Boosting. The study indicates that Ran- dom Forest is able to provide the best accuracy of 89.2% in classifying wine quality, and the most im- portant determinants were alcohol content and volatile acidity. By using extensive feature analysis and model building, we propose the possible relationships be- tween chemical composition and the quality of wine, which may help develop quality measurement devices for wine manufacturing. These results extend the ex- isting library of knowledge based on machine learning and other computational methods in winemaking and suggest that there is a large potential for machine learning to effectively supplement traditional means of wine assessment. Keywords—Wine Quality Prediction, Machine Learn- ing, Random Forest, Feature Engineering, Classifica- tion Algorithms