Predicting Wine Quality from Chemical Properties Using XGBoost Application

Abstract

This research applies XGBoost, a gradient boosting machine learning algorithm, to predict wine quality based on physicochemical properties such as acidity, alcohol content, and sulfur dioxide levels. Traditional sensory evaluations of wine, while critical, are subjective, time-consuming, and prone to variability. By utilizing XGBoost, this study aims to offer a scalable, data-driven approach to automate wine quality assessments, addressing the limitations of traditional methods. The model was fine-tuned through hyperparameter optimization, achieving high prediction accuracy and interpretability. Feature importance analysis provided actionable insights for winemakers, highlighting the key chemical attributes influencing quality. Comparative analysis against Random Forest and Support Vector Machines demonstrated XGBoost's superior efficiency and robustness, particularly in handling non-linear relationships and imbalanced datasets. This research not only enhances the automation of wine quality assessment but also provides valuable knowledge to optimize production processes. The findings underscore the transformative potential of machine learning in the food and beverage industry, enabling consistent quality control and informed decision-making for stakeholders.