Analisis Kualitas Wine Menggunakan Machine Learning dengan Pendekatan SMOTE dan Seleksi Fitur

Conventional wine quality assessment remains reliant on subjective expert judgment, which introduces potential bias and inconsistency in quality control processes. This study aims to develop an objective and automated machine learning-based classification model to enhance the accuracy of wine quality prediction. To address the issue of class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied, along with ANOVA F-test-based feature selection to optimize model performance. The White Wine Quality dataset from the UCI Machine Learning Repository (4,898 samples, 11 numerical features) was utilized to evaluate five classification algorithms: Naïve Bayes, Decision Tree, Random Forest, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN). Before SMOTE application, the Random Forest model achieved an accuracy of only 67.55%. After implementing SMOTE and parameter tuning, the Random Forest (Tuned) model demonstrated the best performance with 90.29% accuracy, 89.99% precision, 90.29% recall, and 89,97%. % F1-score. Additionally, Decision Tree and KNN algorithms also exhibited notable improvements. SMOTE effectively balanced extreme minority class representations (quality levels 3 and 9). The most influential features in quality classification were alcohol content, density, and chlorides. These findings indicate that the proposed framework offers a reliable, objective, and scalable solution for automated wine quality control in industrial production environments.